WIP: CF convention parser

[jbusecke]

• Closes parse CF-style "bounds" coordinates into xgcm axes #127
• Tests added
• Passes pre-commit run --all-files
• User visible changes (including notable bug fixes) are documented in whats-new.rst
• New functions/methods are listed in api.rst
• Double check dependencies if we decide to require cf-xarray

[jbusecke]

This actually works quite neatly out of the box thanks to cf-xarray.

Using this I can take a CMIP6 dataset:

And get a fully functional xgcm grid with a few lines:

kwargs = cf_parser(ds)
grid = Grid(**kwargs)
grid

<xgcm.Grid>
X Axis (periodic, boundary=None):
  * center   x --> outer
  * outer    x_outer --> center
Y Axis (periodic, boundary=None):
  * center   y --> outer
  * outer    y_outer --> center
Z Axis (periodic, boundary=None):
  * center   lev --> outer
  * outer    lev_outer --> center
T Axis (periodic, boundary=None):
  * center   time --> outer
  * outer    time_outer --> center

The parser function modifies the dataset and then passes it + additional kwargs (at the moment only the coords) to the Grid constructor. We could implement this step automatically in the Grid.__init__ if CF conventions are present.

This design follows the ideas layed out in #566 and might also be a good example for #559 (cc @jatkinson1000).

I think once we have #557 done, we can use this PR to refactor the way metadata conventions are parsed, and then add support for SGRID via #559.

[jbusecke]

@dcherian I am curious if you have some ideas on testing here. This works well with a single dataset, but I figured you have a lot more experience with edge cases re CF stuff.

[jbusecke]

Another general question here would be if we want to make cf-xarray a full or optional dependency. cc @rabernat @dcherian @TomNicholas.

[dcherian]

Hah I think the edge cases are in the CMIP archive :)

[jbusecke]

Hah I think the edge cases are in the CMIP archive :)

LOL

[TomNicholas]

We should type hint this, probably with TypedDict