Refactoring group_abundance.

[grst]

The current group_abundance function is relatively slow.
Also I don't like the name too much.

barplot or stackedbarplot is better. x, y and color are better argument names.

Maybe this could even be merged upstream into scanpy.

See also:
https://sc-dandelion.readthedocs.io/en/master/api.html#module-dandelion.plotting

• sort plot segements (Sort bar segments in pl.group_abundance #186)
• get colors from anndata
• get variables not only from obs, but also airr, X, obsm etc.

[grst]

Related #186
Related #247

[grst]

This is a priority now, as with the new data structure, some poor design choices surface.
But we can do this after merging #327.