return the correct number of rows in exp_fibermap when specifying targetids to read_spectra

[moustakas]

Fixes #2125. No unit tests added yet, because my solution is a bit of a hack, if someone passes the rows argument.

@sbailey @dirkscholte

[moustakas]

Unit tests are failing because my code changes assume that EXP_FIBERMAP always exists, which it does not in these tests. I can make the necessary changes, but some feedback regarding what behavior we want would be helpful first.

[sbailey]

Argh, good catch and thanks for the fix. Apologies for bug. As my penance, I have added a test that would have caught it in the first place, and re-added support for read_spectra on non-coadded files that don't have EXP_FIBERMAP.

i.e. uncoadded spectra do not have an EXP_FIBERMAP HDU (it is optional), while coadded spectra do have it, and read_spectra should work with both.

@moustakas if tests pass and you are satisfied with my updates to this PR, we can merge.

[moustakas]

Excellent, @sbailey, thank you. I tested it on my specific case (which were custom coadds) and it works as expected. Thanks especially for the unit test---I struggle to write good tests but always learn from reading yours!