Better fix for flux drop in substrip96 data for NIRISS SOSS extract_1d

[stscijgbot-jp]

Issue JP-3846 was created on JIRA by Ned Molter:

The linked ticket JP-3745 found an unphysical decrease in the flux of the output spectrum in Order 1 between 1.50 and 1.55 um when the SUBSTRIP96 subarray was used.

The problem seemed to be related to a problem with the extraction aperture for Order 2 and/or the location of the trace profile of Order 2 c.f. that aperture.

The fix provided in that ticket was to simply un-mask the entire Order 2 extraction region.  This seems to be adequate for all test cases so far, but is unsatisfying: we didn't really track down the root of the problem.

The current version may also lead to noisier or slightly incorrect extractions because it doesn't properly weight pixels that are half-in, half-out of the extraction region, and it doesn't set the weight of regions that definitely do not contain data to zero.

spacetelescope/pysiaf#362 may also be related.

Link to the lines where the original fix from JP-3745 was inserted is here

jwst/jwst/extract_1d/soss_extract/soss_extract.py

Line 1098 in 8d84c0c

if subarray == 'SUBSTRIP96':