feat(test): add stress-testing to known flaky tests reruns

[JoshuaKGoldberg]

#111860 added a way to rerun known flaky tests 50x locally or in CI. But that didn't reliably reproduce errors on all of them. I think that's because many tests only failed when the CI machine is separately slow, so if the machine is running fine, running 1x vs 50x doesn't make a difference.

This changes flaky tests to run under each of 5 different simulated slowdowns, 10x each (so still 50x each total):

• withDelayedFetch: Simulates a slow network by adding delays to global fetch calls.
• withMainThreadCpuLoad: Simulates an overloaded main thread by doing busy-work on a fixed interval.
• withMemoryPressure: Simulates low available memory by retaining a large buffer of bytes data.
• withMicrotaskChurn: Simulates a busy machine by continuously queueing microtasks.
• withRealWallClockDelay: Simulates a slow machine with many timers by continuously running no-op async delays.

We can see in this PR's CI that although the flaky tests still don't always fail, failures do happen a lot more now. 🥳

Any CI failures (e.g. Jest (1), Jest (2)) are IMO expected and shouldn't block review. The tests themselves aren't changed, and master doesn't run with the flake label.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 4b4de37. Configure here.}