[RM-7100] pdf and probabilities for stars

Created on 2022-03-07 10:27:52 by Didier Vibert. % Done: 0

currently the pdf for stars, build from parameter pdfcombination="bestchi", is constructed by taking at each z the best chi2 among all templates

instead it should be : the pdf build from the chi2 of the same template at all z, being the one that gives the minimum chi2 over all z and all templates.

at the same time we should compute the modelPDF, ie the probability of each template, build from summing the zPDF over z.

and start a thinking about a ML reliability with modelPDF which is scheduled in the Euclid datamodel