Better docs for stopping immediately after failure

[brandon-leapyear]

I'm writing integration tests that I'd like randomized, but don't particular care about being exhaustive, because the tests can take a few seconds to run. So I set max_examples to 1 to stop after the first success, and phases to [Phase.generate] to avoid shrinking after a failure. But I noticed that hypothesis still kept rerunning the tests; bumping the verbosity shows multiple "Trying example" logs, which didn't match the expected behavior I thought hypothesis did.

After scouring the docs, I found the report_multiple_bugs setting, and after disabling it, it seems to correctly stop after the first failure. I think my first point of confusion is why report_multiple_bugs is True by default, as it's not the default behavior for other frameworks like Haskell's QuickCheck or Javascript's fast-check. But separately, it seems to me that there is a hidden phase here, in addition to shrinking, that retries runs to find bugs, that should be possible to be disabled, just like the shrinking phase

[Zac-HD]

I think it makes sense to stop generating immediately on failure if shrinking is disabled, even if report_multiple_failures=True, to better match user expectations.