Deprecate use of assume or reject outside of tests #3748

[nickcollins]

If some code outside a test (e.g. execute_example) calls assume or reject, they raise UnsatisfiedAssumption, which is an internal exception that should generally not seen by user code. This branch provides a warning deprecating such use cases, which helps clarify to the user that they are getting the exception because they have called these functions from the wrong places.

A continuation of #3748 because I messed up the branch on that one and had to create a new branch.

Fixes #3743 and fixes #3753.

[nickcollins]

Some of the tests are failing, including at least one of the tests I wrote, but I don't know how to read the test output to see what is wrong. Is there a way I can run the tests without adding a commit and invoking the CI?

[Zac-HD]

You can click the "details" link next to the checks below this comment, which will lead e.g. here. You can run them locally with e.g. make format or make check-py310-cover; more details here.

[Zac-HD]

Suggested change

	Deprecate use of :func:`~hypothesis.control.assume` and
	:func:`~hypothesis.control.reject` outside of tests.
	Deprecate use of :func:`~hypothesis.assume` and :func:`~hypothesis.reject` from
	outside a Hypothesis context, because these functions work by raising a special exception.

important to note that it's not "outside a test", but outside a particular kind of test. I think it's worth explaining why, too, and would do both in the deprecation message.

[Zac-HD]

...this has turned out to trigger some complicated latent issues in our engine, which are way more complicated than I anticipated. I'll push a commit to fix that, and then merge 🚀

[nickcollins]

Ok, thanks!