Apply spectral smoothing for MRS master background subtraction

[stscijgbot-jp]

Issue JP-3758 was created on JIRA by David Law:

MRS master background currently averages over the spatial area within each wavelength plane, but this still introduces significant noise into the data as the background signal is often dominated by detector noise.  Better results may be obtained by enforcing a degree of spectral smoothness in the master background estimated from dedicated sky observations.

Assignment to MIRI team to study.

[stscijgbot-jp]

Comment by David Law on JIRA:

Kirsten Larson I've drafted a version of this that should work for the MRS, which will likely end up getting handled via https://jira.stsci.edu/browse/JP-3746 that's looking at making other updates for NIRSpec MOS backgrounds (as MOS may use some version of this smoothing as well).

This works by applying a rolling boxcar median to the master sky spectrum that seems to work pretty well at short wavelengths.

One caveat though is that we should not enable it for MRS until after the work has been done to add the dichroic fringe component into the fringe flats.  Otherwise, it will tend to smooth out these dichroic fringes at long wavelengths and thus result in WORSE sky subtraction than before.  I.e., we can get the code in place now, but shouldn't turn it on via parameter reference file until the fringe flats are delivered.