Nightqa: kibo fixes

[araichoor]

This PR should address:

• nightqa fails in some early data when TILEID isn't defined in data header #2356: ignore SCIENCE exposure with no associated fiberassign file (kind of a bugfix, actually...);
• nightqa crashes because it can't find a tile in surveyops tiles-*.ecsv file #2357: ignore tileids for a given survey if it s not in tiles-{survey}.ecsv;
• Nightqa crashes because of "special" survey tile #2358: bugfix (in create_petal_nz()) to restrict to sv1, sv2, sv3, main in the case where all the tileids should be ignored.

Outputs of the desi_night_qa runs for those three cases / nights are here: https://data.desi.lbl.gov/desi/users/raichoor/nightqa_dev/nightqa_v28/

For 20201214 and 20210917, the petalnz-{night}.pdf file is empty, as no tiles are kept for the analysis.

I didn t run desi_night_qa for other nights; hopefully the changes are simple enough that they don t require additional testing... though, I m happy to do so, if you d like me to.

ps: @akremin: sorry for all those issues in night_qa!

[akremin]

Thank you, Anand. No need for an apology. These issues are all on very old data that your code was never designed to run on. I'm glad the fixes to make it robust were straightforward and not too invasive.

I have tested this branch on kibo for nights 20210223, 20210511, and 20210917; which previously failed but now succeed in this branch.