Add cross night information to zdashboard

[akremin]

This updates the redshift dashboard to account for cross night information in the processing of a cumulative tile. This is important when on an earlier night e.g. 10 petals were available but on a later night e.g. 8 petals were available. Before the dashboard would show purple "overfull" values for such tiles because it was expecting 8 redrock files from the 8 petals on that night. Now it correctly identifies that there is petal information for all 10 petals available to the cumulative coadd from the previous night and expects 10 redrock files to be generated.

I tested this on iron, jura, and daily. It worked on all three. Iron shows some missing pernight files, which upon further investigation were accidentally run and then deleted. The dashboard correctly identified them in the processing_table but the state of iron is correct because we didn't want those files.

Jura looks great except for one special tile that we were already aware of that had issues being processed.

Daily looks good for the past two month. Months earlier than that indicate occasional issues in generating a single tile-qa or emline file. Both of these are known issues when coadding very old processings of exposures with modern daily exposures. This only occurs in daily.

Testing code is here: /global/cfs/cdirs/desi/users/kremin/PRs/crossnight_zdash

The dashboards are here: https://data.desi.lbl.gov/desi/users/kremin/PRs/crossnight_zdash/daily_dashboard/zdashboard.html https://data.desi.lbl.gov/desi/users/kremin/PRs/crossnight_zdash/iron_dashboard/zdashboard.html https://data.desi.lbl.gov/desi/users/kremin/PRs/crossnight_zdash/jura_dashboard/zdashboard.html

From these three I don't see any issues with the new cross-night camword logic of the updated dashboard code.

[akremin]

Thank you for the review. I have made the requested changes. The tests are failing for what appears to be an unrelated reason: "ModuleNotFoundError: No module named 'numpy._typing'" in many of the tests.

[sbailey]

It looks like the tests are using scipy 1.14.0 which was released a few days ago, and that uses numpy._typing which doesn't exist in earlier versions of numpy (but maybe does in numpy 2.x which we aren't using?).

I'm not going to take any merge action on a Friday afternoon, but I'm also considering if we should merge this into "jura" instead of "main". This plus PR #2284 would make the final 0.63.x tag to go with Jura, then we merge that branch into main and proceed with other development.

[sbailey]

For the record:

• I constrained scipy<1.14 in requirements.txt to solve the scipy/numpy version incompatibility. We can unpin that in the future when we port ourselves to numpy 2.x and/or the community sorts out the scipy/numpy incompatibility (oddly I didn't find a ticket or other chatter about it).
• Rather than rebase off of jura and then merge into jura, I'll leave this as-is merging into main. That means the final jura tag won't have this change used to generate the final jura ops dashboards, but I think that is ok.