Switch to use romanisim

[yuedong0607]

This RP migrates the codebase to a new set of APIs in romanisim (here's an introduction of the new API), replacing all usages of galsim.roman. This change enables access to both CRDS and the roman-technical-information repository.

Current status:

• Replaced all galsim.roman calls with romanisim APIs.
• Updated the bandpass class to support spatially varying (per-SCA) throughputs, calculated from effective area files retrieved from roman-technical repo.
• Added configurable options for individual instrumental effects in noise.py

TODO / Open Questions (for this PR):

• Configuration schema: Finalize the format of configurable options in the YAML config. Some modules can take additional parameters for further configuration (see an example below for the nonlinearity module)

TODO / Open Questions (out of scope for this PR):

• WCS: Currently still using the same implementation inherited from galsim.roman. Need to consider switching to use the romanisim.wcs APIs which is based on gwcs and leverages the CRDS distortion models.
• PSF: Currently still using the same implementation inherited from galsim.roman. Need to consider switching to use the romanisim.psf APIs which supports both empirical PSF samples from CRDS, and STPSF models. However, both the empirical PSF and STPSF in romanisim.psf only support achromatic modeling at this moment.

from romanisim.models import Nonlinearity
nl = Nonlinearity(
        usecrds=True,
        reftype="inverselinearity",
        getdq=True,
        integralnonlinearity=None,
        metadata=None,
        image_mod=None,
        reffiles=None,
        saturation=None,
        Coeffs=None,
        gain=None,
)
nl.apply(img, electrons=False)

integralnonlinearity : object or None, optional
        Optional integral nonlinearity reference model. If provided, the class
        constructs channel-based lookup corrections that are added after the
        polynomial correction.

[yuedong0607]

romanisim.models has not yet been included in the current release (v0.13.1) of romanisim. The merge needs to wait until the next release version.

[yuedong0607]

I updated the romanisim dependency in pyproject.toml to point to the main branch, and all unit tests are now passing. I think the code is ready for review now. @arunkannawadi

[arunkannawadi]

I have reviewed everything expect for the changes in psf.py. Given the slowness you report, I think that needs some investigation.

[yuedong0607]

I have reviewed everything expect for the changes in psf.py. Given the slowness you report, I think that needs some investigation.

It seems like the efficiency issue appears in both this branch and the integrated_coadd_sim branch (which doesn't use romanisim.models yet). I'm running some tests now.