algorithmic developments in preparation for Fuji

[moustakas]

This PR contains a first round of development work in preparation for Fuji.

Other than some minor clean-up, the most significant issues or improvements addressed so far are:

• fit for the doublet ratio as a free parameter (with physical priors) #39
• Broad+Narrow Balmer line components #47
• Unreal fits because the fit does not converge to a final solution #52

[moustakas]

@Ragadeepika-Pucha @stephjuneau I ran tiles 80605, 80606, 80607, and 80613 with this branch and you can find the full results in /global/cscratch1/sd/ioannis/fuji-dev/everest-v1. (In detail, the merged catalog /global/cscratch1/sd/ioannis/fuji-dev/everest-v1/catalogs/fastspec-everest-cumulative.fits may be all you need.)

There are definitely some outstanding issues but I think this is good progress and I'd be grateful for whatever feedback you may have. Thanks.