Refactor hexbin().

[anntzer]

• Avoid having to copy() x and y, by not overwriting the original
  x and y variables but instead storing the transformed values in
  trfx/trfy.
• Directly construct lattice1 and lattice2 as flat arrays (they are
  flattened at the end anyways), which allows using flat indices:
  the C is None case, becomes a simple bincount, the C is not None
  case can use a list-of-lists instead of an object array.
• Factor out the x/y marginals handling into a for-loop, which
  additionally allows inlining coarse_bin.
• Make the factor of 2 between nx and ny clearer (in the for zname...
  loop setup). (See also [Bug]: Hexbin gridsize interpreted differently for x and y #21349 (comment).)
• Construct marginals verts in a vectorized fashion.

Even if we decide to ultimately deprecate marginals, this should help with converting them to a reusable examples entry.

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).

[jklymak]

Are you claiming the factor of two is expected? This PR doesn't seem to change anything....

[anntzer]

I'm not claiming it is expected (my quick guess is that it isn't), I'm just saying that the current implementation has nbins = nx in one place and nbins = 2 * ny in another, and this refactor shows that more clearly.

[timhoffm]

Avoid having to copy() x and y, by not overwriting the original
x and y variables but instead storing the transformed values in
trfx/trfy.

I find the names trfx/trfy not quite helpful. They are longish compared to x/y but still don't read well. Alternatives:

• Use tx / ty or x_/y_ as minimal extensions to clarify that it's not x/y itself.
• Do the copy on demand if you need to log the values.

[anntzer]

tx/ty it is.

[timhoffm]

This only a partial review. There's possibly more, but I didn't have the time to dig into the algorithm yet.

[jklymak]

I'm not sure we needed this refactor - I'd have preferred something that made the whole thing clearer, but...