Aostools

New capabilities:
- `AddColorbar()`: Add a common color bar to existing Facet Plot (8d1bc12)
- `LogPlot()`: Invert y-axis, make it logarithmic, and change the label format ()
- `Anomaly()`: Compute anomaly for a `xarray.Dataset` or `xarray.DataArray()` (892b66fada54193df5bcb0f1b905e4075073c930)
- `CheckSign()` and `StatTest()`: Perform statistical testing along a given dimension (5efbb58)
- `OceanIndex()` adds 'modiki' option (70cbab363d96b4e30f2fd72034d348f62c6350ae)
- `GlobalAvgXr()`: Compute meridional average, using any power of weighted by cosine of latitude. Does not compute zonal mean. (394a96bf7f07bb039927380743c238c845658e57)
- `FindCoordNames()`: Goes through a list of possible coordinate names to determine the names of pressure, latitude and longitude within a given dataset or dataarray. (0085e3fcc0b98e73dc8b2b6c82f6e583d1d417a1)
- Added more stats tests to `StatTest()`.
- `constants` now includes vorticity parameter and cosine of latitude. (d4fab6cc677b44d48b09dd52690d923cff3381f1)

Many more bugfixes. Thanks for all users providing feedback and allowing `aostools` to become better.

Minor changes include:
- Added `ComputeEPfluxDivXr()`, which computes the Eliassen-Palm fluxes and their divergence with `xarray.DataArrays`. This follows [Jucker, ASL (2021](https://doi.org/10.1002/asl.1020).
- Internal use of `constants` module instead of defining the constants in each function.
- Changed PyPI dependencies: Removed Cartopy for easier install via `pip`. Note that this means if you want to use `Projection()`, you have to install Cartopy separately.

v.2.3.1
This minor update fixes a bug where the PyPI distribution wouldn't install.

Version used for M. Jucker (2020): *Scaling of Eliassen-Palm Flux Vectors*, Atmospheric Science Letters, DOI [10.1002/asl.1020](https://doi.org/10.1002/asl.1020).

Started transition to `xarray` data:
- `ComputePsiXr()`, `ComputeVertEddyXr()` (a9519d4)
- `ComputeWstarXr()` (28c46d2)
- `ComputeRefractiveIndexXr()` (3f22071)
- New function `ComputeRossbyWaveSource()` (3a53884)
- New function `Nino()`: Accepts SSTs and computes nino1+2, 3, 4, 3.4, ONI and TNI. (3a53884)


Minor additions and bugfixes.
- Changed function name `ERA2Model()` to `StandardGrid()` (cee2900)
- Add divergence calculation for `ComputeWaveActivityFlux()`.
- Add possibility to use EOF as input to `eof()`: The function then projects data onto the input EOF and computes the corresponding PC (c192080)
- Add `ComputeStreamfunction()` (f0deda4)
- Bugfixes in `ComputeWaveActivityFlux()`
- Add `Cart2Sphere()` which takes arrow components and transforms them for plotting on a sphere. (607a048)
- `constants.py` now contains often used constants. Will be expanded in the future.
- Multi-axes projection figure with `Projection()` (58ee720)
- Converted spaces to tabs for python3 compatibility.

Due to the rapid development of `xarray`'s capabilities, many of the `aostools` functions based on `numpy` arrays are now much easier to perform. The aim is for future v3 to fully use `xarray` DataArrays. This release is therefore the first to require `xarray`, as some of the added functionality makes use of the much simpler way to manipulate arrays.
New Features:
- 545acc1: `ComputeWaveActivityFlux()` computes Takaya & Nakamura 3D wave fluxes. These are equivalent to EP flux (`ComputeEPfluxDiv()`) if reference state is zonally symmetric and then zonal mean of fluxes is taken.
- f484415:
- `PlotEPfluxArrows()` Accepts `ep1, ep2` computed from `ComputeEPfluxDiv()` and a `figure` and `axes`.
- `Projection()` Creates a figure and axes with a world map projection from `cartopy`.
- `ERA2Model()` Swaps longitude from -180,180 to 0,360 degrees and inverts the direction of latitude to plot EP flux arrows with correct scaling.
- 320f3fae403cb5ef41bd3d84e4dff18929af1c9c: `eof()` now accepts multi-dimensional arrays instead of only 1 time and 1 space dimensions. It assumes the first dimension to be time.
Bugfixes:
- 1d2fffd7960d3c9f32921398f41736985b7b4ed4: `ComputeSaturationMixingRatio()` and `ComputeRelativeHumidity()` are now compatible in terms of input arguments.
- 01368d1: `AxRoll()` and therefore all functions depending on it were not python 3 compatible due to the use of `basestring`, which does not exist in python 3. This is fixed now (and the function was also unnecessarily complicated).
- 71ec5a4: `GetWaves()` now uses `y=None` as default instead of earlier `y=[]`. There was also a bug when `y` was not empty and `do_anomaly=False`. This is now corrected.

This release concentrates on two bugfixes:
- generally make the function more robust if working with the Southern Hemisphere (commit 87dfaaeaf5e9ed72a994604130917ed907a4082c)
- pressure units bug in `ComputePsi` (commit 3b5624a929536537bd8691e2d6e29afeab471038)

A few bugfixes, and extended functionality:
- `GetWaves()` and `ComputeEPDiv()` can now receive a list of wave numbers
- `ComputeAnnularMode()` now deals with both hemispheres - either explicitly with `hemi` or implicitly by looking at the sign of latitude inputs
- `eof()` can now be calculated based on removing the mean (`detrend='constant'`) or a linear trend (`detrend='linear'`).