Add GFA-like TSNR template for dark, bright and backup programs

[michaelJwilson]

Add gfa-like templates to compute gfadark, gfabright and gfabackup tsnrs (same signal, differ by poisson term to IVAR, and soon fiberfrac).

desi_compute_tsnr_ensemble --tracer gfa --outdir .

Creates:
./tsnr-ensemble-gfadark.fits

cp and (SVN) symlink at desimodel/data/tsnr/ to
tsnr-ensemble-gfabright.fits
tsnr-ensemble-gfabackup.fits

Runs through afterburner:

export SPECPROD=daily
export DESI_SPECTRO_CALIB=/global/cfs/cdirs/desi/spectro/desi_spectro_calib/trunk/
export DESIMODEL=/global/cscratch1/sd/mjwilson/GFATEMPLATE/desimodel/

desi_tsnr_afterburner --outfile /global/cscratch1/sd/mjwilson/tsnr-ensemble-updates/tsnr-summary.fits --expids 88359,88360,89881 --recompute --nproc 12 --cameras b0

TSNR2_GFADARK_B = 64.087897
TSNR2_GFABRIGHT_B = 60.688630
TSNR2_GFABACKUP_B = 49.803252

[coveralls]

Coverage decreased (-0.04%) to 27.479% when pulling d705ec4 on tsnr-gfatemplates into d73f092 on tsnr-ensemble-updates.

[julienguy]

Looks great. No comment on the code.
Only one change request: replace the 'tracer' name 'GFA' by something different to avoid confusion with the EFFTIME derived from the actual GFA measurements from Aaron.
A suggestion is 'RBAND' (it cannot be 'R' only because of the confusion with the spectrograph camera arm label). Other ideas welcome.

[michaelJwilson]

Excellent. I'm going to suggest, which I'll implement unless there's pushback, is GPB for Guider Pass Band.

I'm also going to set it up to generate three separate templates GPBDARK, GPBBRIGHT, GPBBackup. These will be the same except for the tracer label in the header.

The symlinks don't really work as some downstream code (not much, but e.g. titles to calibrate_template plots), relies on this tracer keyword.