Turn shared_axes, stale_viewlims into {axis_name: value} dicts.

[anntzer]

This means that various places can now iterate over the dicts
and directly support 3D axes: Axes3D doesn't need to override
_request_autoscale_view or _unstale_viewLim or set_anchor anymore, and
can shed most of set_aspect; various subtle points that were missing
before also get fixed (restoring z-axis sharing upon unpickling;
resetting locators and formatters after deleting an Axes3D sharing a
z-axis with another still present Axes3D); etc.

There's just some slightly annoying interaction with ColorbarAxes
vampirizing its inner_ax, but that can be worked around.

PR Summary

PR Checklist

• Has pytest style unit tests (and pytest passes).
• Is Flake 8 compliant (run flake8 on changed files to check).
• New features are documented, with examples if plot related.
• Documentation is sphinx and numpydoc compliant (the docs should build without error).
• Conforms to Matplotlib style conventions (install flake8-docstrings and run flake8 --docstring-convention=all).
• New features have an entry in doc/users/next_whats_new/ (follow instructions in README.rst there).
• API changes documented in doc/api/next_api_changes/ (follow instructions in README.rst there).

[QuLogic]

Also needs some small doc fix.

[timhoffm]

Since we now have a number of getattr(ax, f"{name}axis"), we may take this one step further and also store the Axises in an internal dict and turn the *axis attributes into properties mapping to that dict:

@property
def xaxis(self):
    return self._axises['x']

But this does not have to be part of this PR.

[timhoffm]

Does this mean we break pickling-compatibility between old and implementation? If so this needs an API-change note.

[anntzer]

The pickling representation is explicitly not stable (see __mpl_version__ in Figure.__setstate__), changes fairly regularly without API note (nearly every time we add an attribute to Figure or Axes, we make old pickles unloadable on new versions, as they would be missing that attribute), and I wouldn't want to start pretending that we put any care in making the pickles stable.

[anntzer]

Agreed that making the axis_map the "real" underlying representation may be best, and also agreed that it can wait for another PR :)

[jklymak]

looks helpful to me!