Dark trail in overscan

[julienguy]

Changes

• compute the dark trail correction prior to the overscan subtraction and apply it to the overscan column.
• add option --keep-overscan-cols to desi_preproc for debugging.

Not in this PR but needed to fix r8 issue: update to $DESI_SPECTRO_CALIB/spec/sm2/sm2-r.yaml (committed but not updated at NERSC): add back the parameter DARKTRAILAMPD: 0.00021 that was lost in recent configurations of this CCD (!)

Comparison before and after the change to the dark trails, but with the desi_proc option --keep-overscan-cols to see the effect. With this change, OSTEPD=0.54 instead of 4.5.

desi_preproc -i /global/cfs/cdirs/desi/spectro/data/20220106/00116955/desi-00116955.fits.fz --cam r8 -o preproc-r8-00116955-new-with-overscan.fits --keep-overscan-cols --nodark --nomask --nopixflat

This solves issue #1619 for r8 (now working on z8 and z3...)

[coveralls]

Coverage decreased (-0.002%) to 25.054% when pulling 47d970e on dark_trail_in_overscan into e343458 on master.

[julienguy]

Result on z8 with calib parameters:

  DARKTRAILAMPD: 0.00006
  DARKTRAILWIDTHD: 28.0

(note it's a 6e-5 effect)

applied to 20220126/00120334 with a very bright cosmic ray hit just on the edge of the active CCD region
OSTEP=1.1 instead of 5.2

[julienguy]

Changes committed to $DESI_SPECTRO_CALIB/spec/sm2/sm2-z.yaml

[julienguy]

Application to z3. With:

  DARKTRAILAMPD: 0.00008
  DARKTRAILWIDTHD: 35.0
  DARKTRAILAMPC: 0.00005
  DARKTRAILWIDTHC: 20.0

Results on 20220111/00118181 z3
OSTEPD from 5.47 to 1.38

[julienguy]

This PR solves issue #1619 (need to update DESI_SPECTRO_CALIB to see the effect).