CI: cosmodc2 is consistently failing in cron

[bsipocz]

One of the cosmodc2 queries is regularly ends up in a timeout and thus fail CI. I feel it's more consistently happening in cron, so we should investigate

• find a better spot to trigger the cron jobs
• shuffle the notebooks around so it's not all the jobs hitting that notebook around the same time
• something else

[troyraen]

Overlaps with #144

[troyraen]

I'm in favor of either this:

shuffle the notebooks around so it's not all the jobs hitting that notebook around the same time

or else making the jobs run sequentially instead of in parallel.

Either of those might resolve B in #175. Other ideas there are:

Spread out so one matrix cell runs each day

only run cron on devdeps and a docs build on latest releases, the rest has a low likelihood of changing outside of PRs.

[bsipocz]

Let's be optimistic about #195 that it cut down crons experienced failures here; we can always reopen if that is not the case.