use more of the precomputed PSFs from Lance - spatial, wavelength interpolation

[rmandelb]

PSF code should be updated in place with spatial variation (and potentially caching); consider also downsampling in wavelength direction and other space-saving measures, as we will have to consider how to distribute the precomputed images

[rmandelb]

This includes:

• Reviewing the header and PDF from Lance to make sure we understand the coordinate system
• Interpolation between multiple different positions: to implement in GalSim-Euclidlike
• Use of >1 CCD: need to make sure we understand the indexing of the quadrants and how that relates to the focal plane layout
• Potential wavelength downsampling
• Consider cropping the stored oversampled postage stamps to reduce data volume
• Figuring out how to distribute data

[rmandelb]

Discussion about spatial variation: start simple, considering these options:

1. No interpolation, just assign the PSF based on the CCD quadrant (i.e. all points in the quadrant get the same PSF). There will be small discontinuities but given our stated purpose of the sims, this is probably OK.
2. Interpolate between quadrant centers, use nearest point without interpolation outside of that.
3. Lie to it that the PSFs at quadrant centers are at the CCD corners, then use the Roman interpolation code.

Proposal given other simplifications in the PSF is to do (1). Once implemented, we'll open another issue to remind ourselves of (2) and (3) for later (but they are not needed soon).

[FedericoBerlfein]

Pushed changes in the pull request Full PSF FOV implementation.