Merge pull request #4152 from RosettaCommons/BYachnin/mhc_epitope_scientific_test_favournative_fix
It turns out that the FavorNativeMover will re-weight all scorefunctions that have a non-zero res_type_constraint. As a consequence, the "base" ref2015 scores that were reported by this scientific test actually included the res_type_constraint contributions, even though it shouldn't have.
This PR uses an alternative mover that doesn't touch the scorefunction weights. Because the this is roughly a constant value added to all scores and for testing purposes, only the percentile rank is used, this shouldn't affect the pass/fail status of the test. The correct values will now be shown in the scorefiles and results page.
I also tweaked the cutoffs to improve test stability.
Because this PR only touches the scientific test files, I will only run the mhc_epitope scientific test +/- debug mode.