DESI_SPECTRO_CALIB psfs with overlapping fibers

[sbailey]

The following two PSFs in DESI_SPECTRO_CALIB have overlapping traces for non-bad fibers:

• spec/sm4/psf-sm4-r0-20210912.fits fibers 135,136 overlap; bad/broken fibers are 367
• spec/sm8/psf-sm8-z7-20210912.fits fibers 3804,3805 overlap (304,305 on just this file); known bad/broken fibers are 3520,3587,3634,3914,3987

This used to be a non-fatal error, but now is a fatal error when that PSF is used as a seed for a traceshift.

TODO: find different PSFs from around that time that could be used instead, perhaps from Iron calibnight.

This is a blocking factor for the Jura run.

Useful commands:

plot_fiber_traces -i spec/sm4/psf-sm4-r0-20210912.fits --fibers 125:150
plot_fiber_traces -i spec/sm8/psf-sm8-z7-20210912.fits --fibers 300:325

specex.qa.specex.qa.trace_psf_qa(psffilename, brokenfibers_string)

Wrapper code in desi/users/sjbailey/debug/badcalibpsf/badcalibpsf.py checked all psffile,brokenfiber combinations that appears in $DESI_SPECTRO_CALIB.

[sbailey]

I replaced the bad PSFs with psfnight files from Iron 20210917, after confirming that the Iron files do not have overlapping traces.