Have the global timeout take into account t.timeout() durations

[OhYash]

This PR attempts to allow tests to run beyond the global limit (which is 10s default).
See #2384 for details.

Tests cases and documentation pending until the debounce() wait period can be overridden.

[novemberborn]

Thanks for turning this into a PR!

You're right that debounce() isn't a direct fit for this. It's perhaps also unnecessary now that we can use https://nodejs.org/docs/latest-v12.x/api/timers.html#timers_timeout_refresh. But maybe that's beside the point.

I don't think the timeout has to be accurate. The goal is for the global timeout to not fire when a test is waiting for a longer period. Perhaps we can think of these t.timeout() calls as preventing the global timeout from firing. We can compute the timestamp before which, if it fires, it should simply restart the timer. What do you think?

[OhYash]

I've updated the api.js to skip global timeout from getting fired in case an individual test timeout is defined.

• Updated some bits for existing tests to still pass.

If this approach looks good, I'll proceed with adding fresh tests specific to the new changes.

This still leaves debounce() limitations to be fixed later. So parallel running tests are more likely to succeed even beyond their timeout limits.

[novemberborn]

Thanks @OhYash! I've pushed some small changes (and merged main). CI is a bit choppy now but some runs should pass…

[OhYash]

Thanks @novemberborn for fixing some of the unnecessary changes I've made. I've fixed the CI tests and added one api.js test

[OhYash]

Added a test for stateChange event emitted by notifyTimeoutUpdate

[novemberborn]

Thanks @OhYash!

[OhYash]

Finally! After me stretching it for too long 😅.
Thanks for help, @novemberborn !!

[mikob]

Thank you @OhYash! @novemberborn will this be included in a release soon? Seems like there hasn't been a release since January.

[novemberborn]

@mikob we're working towards AVA 4. I'm hoping to put another prerelease of that out soon.