Kalepy

- Plotting Improvements
- Allow an `origin` argument to be specified for corner plots, placing the location of the triangle at one of `bl` (default), `tl`, `tr`, `br`.

- New `kalepy.sample` submodule for sampling from continuously defined functions.
- Functionality is defined in the `Sample_Grid` class, and the function `sample_grid()` provides a simple API to construct an instance and use it to construct sample points. Accessible directly from top-level of package.
- `sample_grid_proportional()` and `sample_outliers()`

- Allow `covariance` to be manually specified in KDE constructor.
- New `KDE.from_hist()` method for constructing KDEs based on existing distributions (instead of finite points).
- Deprecated CDF functionality removed (for the time being).
- `Triweight` distribution works.
- Significant PDF evaluation speed improvements using numba
- Sampling and evaluation code simplified.
- Use `abc.ABC` base classes
- `Distribution(object)` ==> `_Distribution(abc.ABC)`
- Plotting improvements
- BUG: fix incorrect label in rotate bottom-right panels of corner plots
- Allow `target` lines to be drawn on corner plots using `Corner.target()`
- Add arguments to limit the number of carpet and scatter points drawn

- DOCS: significant expansion of documentation, both docstrings and sphinx (readthedocs.org).
- Allow `kale.plot.Corner` instances to accept externally-created axes.
- Simplify handling of `reflect` arguments.
- Improve bin-edge guessing.

- Complete restructure of `kalepy.plot` submodule, particularly in the API.
- Extensive addition and improvements of documentation, both inline docstrings, and the addition of sphinx docs now available on [kalepy.readthedocs.io](kalepy.readthedocs.io).
- This includes new demo/test notebooks which are automatically incorporated into the `README.md` and sphinx documentation.
- Documentation, testing, and examples are now included for core plotting functionality. More is needed for the base drawing functions (e.g. `draw_contour2d`, `draw_hist1d`, etc)
- `kalepy` paper
- Fixed typos pointed out by JOSS referees.
- Added citation and comparison to `GetDist` package.
- BUG: `weights` was not being passed correctly during resampling (without reflection).
- MAINT: fixed a huge number of deprecation warnings now raised by numpy about operations on jagged arrays.
- Improved functionality of `kale.utils.jshape` and `kale.utils.really1d` functions to accommodate.
- General plotting improvements
- The handling of colors and colormaps: plotting methods will automatically select the next colorcycle color, construct a matching colormap, and synchronize the color of all plot components.
- The handling of quantiles for confidence and contour components: are now handles much more self-consistently and with a simpler API.
- Drawing functions (e.g. `carpet`, `dist1d` and `dist2d`) will load the current, active axes by default.

- Added paper submitted to JOSS

- API:
- Removed `KDE.pdf_grid` method, instead use `KDE.pdf(... grid=True)`.
- `KDE.pdf(...)` just calls `KDE.density(..., probability=True)`
- NOTE: this means that, like `density` the `pdf()` function now returns a (2,) tuple of the evaluation points in addition to the density values!
- The arguments `reflect` and `params` can now be used in tandem.

- `kalepy/`
- `kde.py`
- `KDE`
- `pdf(...)` is now identical to `density(..., probability=True)`
- `pdf_grid()` [DELETED]
- Call `pdf(..., grid=True)` instead.
- `kernels.py`
- `Kernel`
- `density()` <== `pdf`, `_pdf_clear`, and `_pdf_reflect`
- Combined latter functions into single new method.