data_kind: Add more tests to demonstrate the data kind of various data types

[seisman]

Description of proposed changes

Just add more tests to help us understand the current behavior of the data_kind function.

As can be seen, many data types are recognized as "matrix" and data=None is recognized as "vectors". Sometimes, the result data kind is counterintuitive, as first mentioned in #3351, #3369 and #2744.

This PR only focuses on adding tests. Will refactor the data_kind function in separate PRs.

Please let me if there are other data types that should be tested.

[weiji14]

Erm, do we need tests to check that 1-D and 4-D xarray.DataArray are counted as grid?

[seisman]

Not neccessary. These doctests are added to understand the current behavior of the data_kind function.

I feel 1-D xarray.DataArray should be considered as a vector and 4-D xarray.DataArray should not be recognized and we should raise an exception. Will refactor the data_kind function in a separate PR.

[weiji14]

I feel 1-D xarray.DataArray should be considered as a vector

We already recognize an xr.Dataset composed of 1-D xarray.DataArrays as a matrix (can be treated similarly as multiple 'columns' in a Dataframe). I don't think we need to expand this logic, since they users can call .data on a 1-D xarray.DataArray to get a numpy.ndarray that can then be passed to the data parameter.

4-D xarray.DataArray should not be recognized and we should raise an exception

Logic could be something like if dataarray.ndim() >= 4: raise Error, but I don't think this is really necessary since an error would be raised by the PyGMT module anyway right?

[weiji14]

Pre-approving since most of this looks good. There's some discussion about whether to check 1-D/4-D xarray.DataArrays at #3480 (comment), but not very critical and can be handled later with the actual code refactor.