handle missing cframe in tsnr2 summing

[araichoor]

This PR is a simple suggested fix for issue #1664.
I let @julienguy review.

My understanding is that for each fiber of a given petal, we get the TSNR2 with summing the values over the three cameras, i.e.: TSNR2 = TSNR2_ELG_B+TSNR2_ELG_R+TSNR2_ELG_Z (using ELG here)

desispec/py/desispec/exposure_qa.py

Lines 305 to 312 in 769b712

	tsnr2_for_efftime_vals = np.zeros(entries.size)
	for band in ["B","R","Z"] :
	camera="{}{}".format(band.lower(),spectro)
	cframe_filename=findfile('cframe',night,expid,camera,specprod_dir=specprod_dir)
	scores = fitsio.read(cframe_filename,"SCORES")
	tsnr2_for_efftime_vals += scores[tsnr2_for_efftime_key+"_"+band]
	target_type=tsnr2_for_efftime_key.split("_")[1].upper()
	efftime = tsnr2_to_efftime(tsnr2_for_efftime_vals,target_type)

(entries are the 500 indexes of the fibers for the considered petal).

This fix will just ignore a camera if the cframe file is missing.
In the issue case (z-camera missing for petal=7), we will have for petal=7: TSNR2 = TSNR2_ELG_B+TSNR2_ELG_R, and this will result in a lower efftime:

>>> d = Table.read("/global/cfs/cdirs/desi/users/raichoor/tmpdir/exposures/20210406/00083716/exposure-qa-00083716.fits","FIBERQA")
>>> for p in range(10): print("{}\t{:.0f}s".format(p, d["EFFTIME_SPEC"][d["PETAL_LOC"] == p].mean()))
... 
0	698s
1	790s
2	839s
3	690s
4	855s
5	707s
6	963s
7	148s
8	663s
9	723s

Is that what we want?

[julienguy]

yes, thanks. I am merging.