[RM-9207] [powerLaw] implement fastigm processing

Created on 2024-09-17 12:27:14 by Didier Vibert. % Done: 0

like for template fitting, to speed-up the processing, the powerlaw fitting with igm extinction can be optimized: see fastigm processing in templatefitting operator.

overview, 2 cases:

1. at redshifts where Lya (1216 Angstrom) is not observed: no need to loop on igm
2. when 1216 is in the observable range, if there are some samples redder than 1216, those samples are not extincted and thus the model is the same for some portion of the wavelength range. The idea is to not recompute all the sums involving the model in the computation of the linear least-square fitting, by splitting these sums in two: one part with the extincted portion of the model and the other part with the non-extincted portion computed only once. And apply same trick in the chi2 computation.