Adapt Roman code to new required data files

[ojustino]

Cleans up my test branch with code changes to handle the new structure of the required Roman data from GSFC – OPDs, pupils, and expanded Zernike coefficients.

Let's see if changing DATA_VERSION_MIN in stpsf/__init__.py is enough to enforce the use of the new files as we test this branch. If we merge this branch before the release is ready, we may need to revert this change so others who want to use the develop branch but don't have the files can still run STPSF.

I included @mperrin's quick fix to prevent large slowdowns from the number of Zernike coefficients exceeding the lru_cache's default maximum size, but we could still benefit from either aligning the calculation more with the approach taken in STPSF's JWST models (future task) or by making the R() function from poppy's zernike.py more efficient.