Update pinned dependencies

[Zac-HD]

Automatically update pinned dependencies

[Liam-DeVoe]

This release includes the black 2026 style: psf/black#4875

[Liam-DeVoe]

@pschanely looks like an infinite hang in crosshair, new in this set of version bumps: https://github.com/HypothesisWorks/hypothesis/actions/runs/21324520304/job/61389065468?pr=4646

(still going at ~2 hours at time of comment, will probably have timed out by the time you read this)

[Liam-DeVoe]

Not an infinite hang, just very slow:

8504.60s call hypothesis-python/tests/nocover/test_recursive.py::test_can_find_quite_broad_lists

This seems new in the latest crosshair, though it could just be very coincidental timing. I've skipped here.

[Liam-DeVoe]

I sat down for the second time this afternoon and tried to drill down into the perennially-flaky database listener tests. I've concluded there's likely some real issues here - which is why I've held off disabling them up to now - but that fixing it would require investigating at least one level down the stack (into watchdog), and possibly two (into OS filesystem watchers), which I don't have medium-future time for.

Some of my dirty investigations here: Liam-DeVoe#9

I'll try to write up an issue on this so we have a record

[Zac-HD]

Pandas 3.0 makes pytz an optional (and discouraged) dependency, which led me to notice that we were pointlessly installing it for our coverage tests. I also added a note that Django will drop pytz support when django 4.2 [LTS] goes end of life this April; at that point we should probably stop installing it for hypothesis[all] and add a 'upstream library deprecated' warning to our docs.

It's a nice ecosystem-level cleanup to have better timezones standardized now, but I'm not really in a rush to remove the strategy from Hypothesis - maybe in a year or so...

[Liam-DeVoe]

(unsticking..)