MWS colors are oddly gridded in the dc17b truth catalogs

[moustakas]

@segasai has identified odd / gridded colors in the MWS truth catalogs in the dc17b run, suggestive of an indexing bug somewhere. Track this down and fix it.

[moustakas]

@segasai I looked into this issue in some detail as part of a refactor of the code which connects the Galaxia mocks to spectra, and I don't believe this is a bug -- it does indeed appear to be due to the gridding in the stellar templates, i.e., due to the fact that we're not interpolating in Teff and other stellar properties.

For example, here's apparent (DECaLS) g-r color vs Teff for a mini dataset of a couple thousand MWS_MAIN targets:

A high-priority ToDo item is to interpolate the stellar spectra so that this issue can be eliminated -- and @callendeprieto has code to do so -- but I have not had the cycles to incorporate it into the simulation infrastructure.

[moustakas]

I'm fairly confident that this is an understood issue that should have improved somewhat with the higher density of stellar templates added in desihub/desisim-testdata#9 and used in #264. Meanwhile, we're also working on an improved (higher-density, irregularly gridded) set of stellar templates.