Include chi2 of excluded datasets in training reports

Created by: scarlehoff

People in the mailing list have manifested their wish to have the chi2 of excluded datasets in the training reports. I think it could be a nice addition (but I would have it as a separate table).

An example of such a table is given by @enocera here https://vp.nnpdf.science/o7dD198iRsSv_UFS6bwPAw==

Note that this is not trivial since it can only be done in some cases (for instance, if the theory choices are different in both reports, such a table cannot be constructed in an unambiguous way).