Split extra libraries and add more granular control over tests

[Zeitsperre]

Overview

This Pull Request splits out the libraries that support some complex validators (geotiff, netcdf4, GDAL/OGR, zarr) so that basic data types are supported by default.

It also modifies some of the test controls so that pytest markers can be used to explicitly select/deselect tests that rely on these extra libraries.

Finally, coverage is now done across all builds in parallel, using tox-gh and the coverallsapp/github-action to remove the coveralls-python library.

Related Issue / Discussion

Changes here build on discussions found in #698

Closes #698

Additional Information

There are many more changes I'd like to implement, but I'll limit things here for now until discussion is had in #700.

Contribution Agreement

(as per https://github.com/geopython/pywps/blob/master/CONTRIBUTING.rst#contributions-and-licensing)

• I'd like to contribute [feature X|bugfix Y|docs|something else] to PyWPS. I confirm that my contributions to PyWPS will be compatible with the PyWPS license guidelines at the time of contribution.
• I have already previously agreed to the PyWPS Contributions and Licensing Guidelines

[coveralls]

coverage: 80.551% (-3.6%) from 84.104%
when pulling 250b4f5 on extras-requirements
into 04f734b on main.

[Zeitsperre]

@cehbrecht I'm not certain why the coverage status is taking such a huge hit. I did remove a few useless imports, so it could be that the original values were too high?

[fmigneault]

@Zeitsperre
Potentially because the package were installed by default, their tests were running and checking some code paths (like their input format validation for example). Now, they would be skipped if not installed, therefore not "covered".

[Zeitsperre]

Now, they would be skipped if not installed, therefore not "covered".

I have split the builds with tox to report collective coverages based on the sum from lines covered by default dependencies and the extra dependencies. The logic for both configurations (with and without extras) is being hit

There were a few tests that I disabled/skipped because there were no assertions made with the results: smoke tests, essentially. It's likely from those changes that the coverage is being impacted. In any case, I think the coverage score is fine.

[Zeitsperre]

@fmigneault I figured out why the coverage was much higher: Previously, the values were reflective of both the source code (pywps) and the tests. The fact that the values are now restricted to the library, we lost around 4% of average coverage.

The value currently is also impacted by the removal of a handful of smoke tests, like I mentioned, but on the order of ~1%. Getting that back would require request mocking and other tricks. Are you familiar with those approaches?