Refactor axis_physical_types

[Cadair]

This is the second part of #457

The main work here is refactoring the logic of deciding the default coordinate frames from the base class into subclasses. Also add a PixelFrame which supports multiple dimensions. I haven't done anything with Frame2D here, but imo this is a complete replacement for it?

[codecov]

Codecov Report

Patch coverage: 97.77% and project coverage change: -0.35 ⚠️

Comparison is base (8ded81c) 87.25% compared to head (1c20498) 86.90%.

❗ Current head 1c20498 differs from pull request most recent head b61372d. Consider uploading reports for the commit b61372d to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #459      +/-   ##
==========================================
- Coverage   87.25%   86.90%   -0.35%     
==========================================
  Files          23       23              
  Lines        3820     3857      +37     
==========================================
+ Hits         3333     3352      +19     
- Misses        487      505      +18     

Impacted Files	Coverage Δ
gwcs/coordinate_frames.py	95.26% <97.77%> (-0.12%)	⬇️

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[nden]

These changes look good to me overall. It's unclear what the return type should be. In the FITS implementation it's defined as iterable and the examples return a list. GWCS uses tuples. How important is to be consistent т тхат лежел?

The new PixelFrame does not add much, however if it's convenient, I'm fine with adding it. I should note that there are no tests with this PR. And the new frame will need a Converter.
JWST regression tests pass: https://plwishmaster.stsci.edu:8081/job/RT/job/JWST-Developers-Pull-Requests/750/

[Cadair]

I would say tuple rather than a list is perfectly fine, iterable is iterable. 🤷

[Cadair]

I have pulled the PixelFrame commit from this PR, will add it to its own one.

[nden]

Failure is unrelated

[nden]

@Cadair I'm trying to sort out the jwst regression tests and once done will merge this.