Extend to consider multiple stars

[benmontet]

I propose we fit for [Nstar x Ncadence] fluxes, [Nstar x2] positions (x and y), and [(Nstar-1) x2] offsets between stars. As the detector moves around the relative positions of stars will be fixed, so any apparent changes in the centroids of each is actually telling us something about the flat field.