Harmonize cmap handling in map functions

[Beauprel]

This PR introduces changes to harmonize colormap handling in gridmap, scattermap and gdfmap.

The goals were to:

1. Make sure all combinations of the cmap, levels and divergent arguments worked properly in the three functions.
2. Remove the dependency to the IPCC discrete colormap RGB files.

The following changes were made:

• fixed bug causing util custom_cmap_norm to not work when the cmap was supplied as a string.

• vmin, vmax now work in plot_kw of gdfmap

• fixed bug that happened when levels was a list and divergent was an int in gdfmap

• removed dependency on discrete IPCC colormaps from gdfmap

• Removed dependency on discrete IPCC colormaps from gridmap

  • When the colormap is not discrete, custom_cmap_norm works and we just have to feed norm into xr.plot.pcolormesh() like in the other functions
  • However, when the cmap is discrete (levels), because of a bug in xarray (Colormap Normalisation Giving Unexpected/Incorrect Output pydata/xarray#4061 ), a list of levels has to be passed in the levels argument of xarray.plot.pcolormesh( ) . To do so, custom_cmap_norm now returns a list of levels instead of a normalization instance, in certain cases (when levels is not False in gridmap only).
  • Removed the util cbar_ticks, which was useless anyways because xarray already reduces the number of ticks on the cbar when there are too many,
  • Removed the levels argument from create_cmap, and the discrete cmap IPCC data
  • Removed the maximum number of levels in gridmap, which was entirely due to the IPCC files.

[juliettelavoie]

Il restera un peu de travail à faire sur les colorbars comme discuter sur teams, mais je pense que c'est bon pour l'instant !

[juliettelavoie]

change to docstring accordingly ( for all functions)

[Beauprel]

Maybe a possibility to quickly make levels nicer: change the dtype of the linspace calls in custom_cmap_norm, when the data range is big enough.

e.g. np.linspace(1,10, 10, dtype=int)

This could result in uneven intervals, but you can use plt.colorbar(spacing='uniform') to make them all appear the same on the colorbar.