New hierarchy for Mapping classes

[Sworzzy]

This PR is to link with Sympde PR#168, where the new hierarchy is showed.

The interface is now simpler. An example can be in the files changed of this PR, in psydac/api/postprocessing.py, in the _get_mesh function l. 2286 :
There is no callable mapping anymore. So the type of the mapping now is either : spline, analytic or None. ANd you may do directly evaluations on the object without getting a callable argument or so.

This new hierarchy maintains "domain undefined mapping", that you may get from the Domain.from_file method.
Now, in the tests, the real mappings (the "defined" or "callable" version), can be obtained from domain_h.mappings (returns a list of mappings, either spline or analytic) : see psydac.api.tests.test_api_feec_2d.py

An improvement could be done in this mapping encapsulation. Which class should have the "defined" mapping, to have the most user-friendly tests. For example in psydac.api.tests.test_api_feec_2d.py, you must set the mapping attribute, of the derham object :

if use_spline_mapping:
        domain_h = discretize(domain, filename=filename, comm=MPI.COMM_WORLD) 
        derham_h = discretize(derham, domain_h, multiplicity = [mult, mult])
        
        # TO FIX : the way domain.from_file and discretize_domain runs make that only domain_h.domain.interior has an actual 
         SplineMapping. The rest are BaseMapping.
        # The trick is to now when to set exactly the BaseMapping to the SplineMapping. We could try in discretize_derham, but 
        see if it doesn't generate any issues for other tests.

        mappings=list(domain_h.mappings.values())
        
        if(len(mappings)>1):
            raise TypeError("we are not doing multipatch here")
        
        mapping = mappings[0] # mapping is SplineMapping now
        derham_h.mapping=mapping

[yguclu]

Great job @Sworzzy!

You have erronously included some macOS hidden files in the PR

[Sworzzy]

I've deleted all the .DS_Store files and modified the .gitignore suscessfully.

[yguclu]

Hey @Sworzzy, good effort! I am posting a partial review.

[campospinto]

the CollelaMapping2D from sympde has k1 and k2 as parameters. This line is probably wrong.