Also scan for large trace shifts in continuum/LED exposures

[julienguy]

Second PR on trace shifts, this time to address the fit of continuum/LED exposures when there are large trace shifts
(the previous PR was for arc lamp exposures).

• Add a first scan for large trace shifts when running desi_compute_trace_shift with option --continuum.
• Add option --no-large-shift-scan to disable this feature.
• This feature is not used in daily pipeline reduction of flat field exposures because a correct PSF (fitted from previous exposures is used).

Example:

desi_preproc -i /global/cfs/cdirs/desi/spectro/data/20250214/00279014/desi-00279014.fits.fz --cam r7 -o preproc-r7-00279014.fits

desi_compute_trace_shifts --psf /global/cfs/cdirs/desi/users/jguy/teststand/desi_spectro_calib/spec/sm8/psfnight-r7-20250120.fits --continuum -i preproc-r7-00279014.fits --no-large-shift-scan -o psf-r7-00279014-no-large-shift-scan.fits
...
INFO:trace_shifts.py:697:fit_trace_shifts: MEANDX = -1.44 pixel
INFO:trace_shifts.py:698:fit_trace_shifts: MEANDY = +0.00 pixel

desi_compute_trace_shifts --psf /global/cfs/cdirs/desi/users/jguy/teststand/desi_spectro_calib/spec/sm8/psfnight-r7-20250120.fits --continuum -i preproc-r7-00279014.fits -o psf-r7-00279014-new.fits
...
INFO:trace_shifts.py:697:fit_trace_shifts: MEANDX = +5.80 pixel
INFO:trace_shifts.py:698:fit_trace_shifts: MEANDY = +0.00 pixel

Without the large shift scan:

plot_fiber_traces -i psf-r7-no-large-shift-scan.fits --image  preproc-r7-00279014.fits  --vmax 1000

With the large shift scan (default when running with option --continuum)

plot_fiber_traces -i psf-r7-00279014-new.fits --image  preproc-r7-00279014.fits  --vmax 1000

[julienguy]

Note that this PR does not affect the daily processing.

[coveralls]

coverage: 38.94% (-0.05%) from 38.986%
when pulling 253f251 on traceshift2025b
into c717572 on main.

[rongpu]

Some brief comments where large trace shift is triggered would be useful. This PR looks good otherwise.