Stop emitting misleading warning when constructing results of a fit with daemon_flux parameters

[thehrh]

When performing the sanity check on the metric value: Summing all the individual prior contributions was wrong, and so was the associated warning.

The detailed metric info should have been extended by the correct calculation by ParamSet.priors_penalty instead, which it now includes and uses (if available: old fit results won't have the "priors_penalty_total" entry/entries but still need to be supported—though without any comparison between metric values now).

In the Detectors case, each DistributionMaker's detailed metric info will also contain the new entry, but we still can't just add these up due to the possible existence of shared parameters.

No unit tests are added here, but I've tested the following manually:

• at the end of a fit using a Pipeline with or without free daemon_flux-parameters (I modified IceCube_3y_oscillations_example.ipynb for this):
  • no warning is emitted
  • result.detailed_metric_info["mod_chi2"]["priors_penalty_total"] indeed contains a value that when added to result.detailed_metric_info["mod_chi2"]["maps"]["total"] agrees with result.metric_val
• in the case of the Detectors fit in Multi_detector_test.ipynb:
  • as expected, there are two identical entries for "priors_penalty_total"
  • also: 2 * result.detailed_metric_info[0]["mod_chi2"]["maps"]["total"] + result.detailed_metric_info[0]["mod_chi2"]["priors_penalty_total"] yields the same value as result.metric_val
    (The first expression isn't generally valid, just in the special notebook scenario.)