We want to be able to do a whole-spectrum fit, rather than row-wise slices, so a PSF model is required.

We should be able to support some analytical/physical model, and also doing something arbitrary as the lab-PSF is not normal (some kind of flat-topped Gaussian maybe?)