RCA-1039: refactor roman_catalog_process.py to only run informer stage when necessary.

[mairanteodoro]

Resolves RCAL-1039

The refactoring of roman_catalog_process.py was intended to prevent running the informer stage unnecessarily. It will check for the existence of a previously created model that contains all the information necessary for the estimation stage (the grid of models), which is where we actually determine the best model. roman_photoz will now check for the presence of a pickle file named "roman_model.pkl" (default) or with the name provided by the user through the --model_filename argument. If the file exists, we'll use it, otherwise, we'll run the informer stage to create one. In either case, the file location is in the $LEPHAREWORK dir.

[schlafly]

This looks good. Can you confirm that if only that pickle file exists, the "estimator" stage would run successfully?

If something like that is true, we should consider putting that pickle file in CRDS and updating it periodically. And temporarily putting that pickle file somewhere like /grp/roman/scsb/roman_photoz_data/roman_model.pkl
where it would get picked up. And then the default mode of operation would be for it to look at a LEPHARE_MODEL environment variable or something to pick up that file; we'd create it with a separate mechanism. Does something like that make sense to you?

[mairanteodoro]

This looks good. Can you confirm that if only that pickle file exists, the "estimator" stage would run successfully?

Yes, there is a unit test for that in this PR (test_process_model_existence()).