Implement Sara Profile (Drude convolution)

[jdtsmith]

@sarduval has explored Gaussian x Drude convolutions (the functional form of which we nickname "the Sara profile". She's come up with a mapping scheme to update a given Drude to another broader Drude, which works for most reasonable ranges of resolution and PAH features. Right now our "parent" model for Drudes bakes in the instrument line spread function convolution: not good.

The idea is that science packs should be "pure" and devoid of any artifacts of the instrument. So just as we express lines as Gaussians of a computed FWHM, we'd compute updated FWHM for Drude's too. Other features (BB, modifiedBB) are too broad for this to matter much.

@drvdputt, any thoughts on where this should go? I guess the Fitter is in charge of marrying instrument pack, figuring out gaussians, etc.? So it should do this too. Do we update fwhm for lines in the Features table after a fit? We need a way of taking a fitted FWHM and running the transform backwards: so we can recover the true intrinsic feature width.

[jdtsmith]

Relevant comment from PR#289, regarding redshift:

When we expand Drude profile widths ala #288, we'll need to do something similar with the teleported line profile width, i.e. providing "the width of the line profile at the observed center of the feature, cast into the rest frame". We should probably clearly spell this out in a comment here.