Kuibit

`SimDir` can now be cached in pickle files

`kuibit` tries to do as much lazy-loading as possible. For examples, files are
opened only when needed. When analyzing simulations it is useful to save the
work done by `kuibit` to avoid re-doing the same operations over and over. It is
now possible to do this using pickle files. `SimDir` can now be used as a
context manager and the progresses can be loaded and saved from files. For
example:

python
with SimDir("path_of_simulation", pickle_file="simdir.pickle") as sim:
do operations


In this case, if `pickle_file` exists, it will be loaded (ignoring all the other
arguments passed to `SimDir`), and it will be kept updated with the additional
work done by `kuibit`. If `pickle_file` does not exist, the `SimDir` will be
created as usual as a `pickle_file` will be generated.

It is important to stress that, when using pickles, no consistency check with
the current state of the simulation is performed. If the simulation changes
(e.g., new checkpoints are added), this will result in errors. In that case, a
new pickle file must be produced.

Masked data

Numerical objects now support
[mask](https://numpy.org/doc/stable/reference/maskedarray.generic.html), which
can be used to ignore part of the data that satisfy certain conditions (for
example, to exclude the atmosphere from GRMHD computations).

Note that it is not possible to perform interpolation with masked data, so
several methods will not work.

Note also that we mask only the data, not the independent coordinate (e.g., the
time in TimeSeries or the spatial coordinates in the UniformGridData). In case
you need masked coordinates too, you can use the `mask` method to obtain an
array of booleans that identifies the valid data.

- `Series`, `UniformGridData`, `HierarhicalGridData` have a new method
`is_masked`.
- `Series`, `UniformGridData`, `HierarhicalGridData` have a new method
`mask` that identifies where data is invalid.
- `Series`, `UniformGridData`, `HierarhicalGridData` have news methods to create
masked data (e.g., `mask_greater`). See complete list in documentation.
- `Series` have new methods `mask_remove` and `mask_removed` to create objects
without masked data.
- `Series` have new methods `mask_apply` and `mask_applyed` to create objects
with a given mask (as the one obtained with the `mask` method).

General
- `SimDir` can be saved to disk with the method `save` and read with the
function `load_SimDir`. This is useful to work with a simulation that has
finished.
- Examples can now use pickles.
- New CI workflow: linting.
- New `First Steps` documentation page.
- When a release is published, its documentation is saved to
`sbozzolo.github.io/kuibit/VERSION`

Features

- `time_at_maximum` and `time_at_minimum` in `TimeSeries` can now take the
optional argument `absolute`.
- Added `x_at_minimum_y` and `x_at_maximum_y` to `BaseNumerical`.
- Added `coordinates_at_maximum` and `coordinates_at_minimum` for grid data.
- Added `HierarchicalGridData.is_complex()`.
- Added `tikz_clean_figure` to `visualize_matplotlib.save`, to
`argparse_helper.add_figure`, and to examples. This can be used to reduce
the size of output `tikz` files.
- Added `clear_cache` in `OneGridFunction`.
- Added `plot_contour`.
- Added alias `time_found` to `formation_time` in horizons.
- Added `plot_components_boundaries`.
- Added `ghost_zones_remove` in `HierarchicalGridData`
- Added `add_grid_structure_to_parser`.
- Added `reflection_symmetry_undone`.

Breaking changes
- The `ignore` parameter in `SimDir` has been renamed to `ignored_dirs`.
- The `trim_ends` parameter in `cactus_waves` is now set to `False` by default.

Bug fixes

- `plot_colorbar` does not steal axis focus anymore.
- The legend in `plot_psi4_lm` was corrected.
- `visualize_matplotlib.save` now correctly supports the `figure` argument.
- `plot_strain_lm.py` no longer crashes when `window_args` is not provided.
- `HierarchicalGridData` now owns the components.
- Clear `OneGridFunction` cache in `grid_var` to avoid death by OOM.
- Uniform constructor of `GridSeries` with constructors of other `Series`.
- Horizon properties now lead to valid python variable names.

New examples

Examples with `--detector-num` now also accept `--num-detector` as alias.

Scripts:

* `picklify.py`
* `plot_1d_slice.py`
* `plot_grid_expr.py`
* `plot_phi_time_averaged.py`
* `plot_gw_angular_momentum.py`
* `print_grid_point_minmax.py`

New module: visualize_matplotlib

The `visualize_matplotlib` module aims to simplify common visualization tasks
with `matplotlib`. At the moment, it mainly supports visualizing grid data and
apparent horizons outlines. The public functions in `visualize_matplotlib` try
to be as general as possible: if you pass some grid objects, they will try to
figure out how to plot it. Nonetheless, you should read the documentation and
the docstrings of the various functions.

`motionpicture`

Making movies is a critical step in analyzing a simulation. Now, `kuibit` comes
with [motionpicture](https://github.com/Sbozzolo/motionpicture), a Python tool
to assist you animate your data. `motionpicture` provides all the infrastructure
needed to render multiple frames and glue them together, so, all you need to
worry is how to render one single frame. Importantly, `motionpicture` supports
parallel rendering, which can dramatically speed up the time needed to produce a
video. Check out the
[grid_var](https://github.com/Sbozzolo/kuibit/blob/master/examples/mopi_movies/grid_var)
example to see how easy it is to make a movie.

New class: GridSeries

`GridSeries` is a new class to describe 1D grid data. This class utilizes the
same infrastructure used by `TimeSeries` and `FrequencySeries` to represent a
single-valued function. `UniformGridData` can be transformed into `GridSeries`
with the method `to_GridSeries`. The main reason you would want to do this is
because `GridSeries` are leaner and more direct to use.

General
- Improvements to documentation, docstrings, and tutorials
- Examples are now automatically packaged and uploaded upon release
- New YouTube series, `Using kuibit`

New examples

Scripts:

* `plot_1d_vars.py`
* `plot_ah_coordinate_velocity.py`
* `plot_ah_found.py`
* `plot_ah_radius.py`
* `plot_ah_separation.py`
* `plot_constraints.py`
* `plot_em_energy.py`
* `plot_grid_var.py`
* `plot_gw_energy.py`
* `plot_gw_linear_momentum.py`
* `plot_physical_time_per_hour.py`
* `plot_phi_lm.py`
* `plot_psi4_lm.py`
* `plot_strain_lm.py`
* `plot_timeseries.py`
* `plot_total_luminosity.py`
* `print_qlm_properties_at_time.py`

Movies:

* `grid_var`

Bug fixes
- Fixed header recognition for `carpet-grid.asc` (22)

Features

- New methods `get_apparent_horizon` and `get_qlm_horizon` in `HorizonsDir`.

Bug fixes
- Fixed corner in `__str__` in `UniformGrid`
- Fixed `_finest_component_at_point_mapping` for some points for which floor and
rounding lead to different results.

New module: argparse_helper

The `argparse_helper` module collects functions to set up arguments for
command-line scripts. It comes with options for working with figure, grid data,
and horizons. Options can be stored in text files and read passing the path to
the `-c` flag.

Faster `HierarchicalGridData`

The function that finds the component corresponding to a given point in
`HierarchicalGridData` was significantly sped up by adopting the algorithm used
in `PostCactus` (developed by Wolfgang Kastaun). This new algorithm is
significantly faster, but it will not work in case the refinement factors across
different levels are not constant integers. In that case (e.g., the refinement
boundaries are at 1, 2, and 6), the older algorithm will be used. For large
towers of refinement levels and hundreds of MPI processes, the new algorithm is
orders of magnitude faster.

Examples

Now `kuibit` comes with runnable examples. These are production-grade codes that
you can immediately use for your simulations. They are a great way to learn about
how to use `kuibit`. The examples included are:

* `print_ah_formation_time.py`
* `print_available_iterations.py`
* `print_available_timeseries.py`
* `save_resampled_grid_data.py`

General

- Releases are now automatically pushed to PyPI.

Features
- Added method to compute the linear momentum lost by gravitational waves
- Now the method `save` in `UniformGridData` supports `.npz` files. This is the
recommend and fastest way to save a `UniformGridData` to disk.
- The function `load_UniformGridData` can now read `.npz` files.

Bug fixes
- Fixed bug that, under certain circumstances, resulted in `cactus_grid_function`
not correctly indexing all the 3D data files
- Fixed a test that was triggering the wrong error.
- Fixed a bug that made `ra_dec_to_theta_phi` depend on the local time

Breaking changes
- `finest_level_component_at_point` is now `finest_component_at_point` and
returns directly the component as `UniformGridData` as opposed to the
refinement level and component number.

General
- Improvements to documentation and docstrings

Features
- Added `nanmax`, `nanmix`, `abs_nanmax`, `abs_nanmin` to `BaseNumerical`
- Added support for HDF5 grid arrays
- Added slicing `HierarchicalGridData`
- Added `shape_time_at_iteration` in `cactus_horizon`
- Added `shape_at_time` in `cactus_horizon`
- Added `shape_outline_at_time` in `cactus_horizon`
- Added `compute_horizons_separation` in `cactus_horizon`
- Added `ignore_symlinks` to `SimDir`

Bug fixes
- `cactus_horizons` and `cactus_multiploes` now remove duplicate iterations

Breaking changes
- `remove_duplicate_iters` was renamed to `remove_duplicated_iters`
- `_derive` methods are renamed to`_differentiate`

Initial release