Shower/snowball detection in the likelihood-based ramp fitting

[stscijgbot-jp]

Issue JP-3767 was created on JIRA by David Law:

Timothy Brandt and I got to spend some time discussing how to integrate shower and snowball flagging with the likelihood-based ramp fitting algorithm.  While this ramp fitting provides better shower/snowball suppression than NOT using the dedicated algorithms, it isn't as good as those algorithms.

Skipping straight to the resolution, we determined that the likelihood-based approach can simply use the jump flags already determined by the JUMP step.  I.e., regular jump step should always be run regardless of what algorithm is used for ramp fitting (i.e., as is already necessary for the 1/f correction step).  The likelihood-based ramp fitting is only special inasmuch as it revises the jump flagging further during the ramp fitting itself.

This may require some changes to the likelihood-based code, but such changes are ON HOLD pending further investigation and testing of https://jira.stsci.edu/browse/JP-3765