Ramp fitting multiprocessing regression test revealed read noise variance bug

[stscijgbot-jp]

Issue JP-3791 was created on JIRA by Kenneth MacDonald:

The creation of a multiprocessing regression test for ramp fitting revealed a bug in computing the read noise variance for particular conditions.  Specifically, I think this is related to the CHARGELOSS re-computation of the read noise variance after fitting the slope.  All of the failing pixels that have been manually examined show some CHARGELOSS flagging.  Also, I can't think of any other type of bug that would impact read noise variance and nothing else.