regr_test.py: Run mypy with --no-incremental

[AlexWaygood]

Fixes #11220. Closes #11344; closes #11277. I'm currently stress-testing this locally to see if it does actually fix the intermittent crash. Creating a PR as well so that we can see if this significantly slows things down at all for us in CI. If it does, we'll be better off going with something like #11344, regardless of whether this fixes the crash.

[AlexWaygood]

The good news is that this doesn't seem to slow things down much in CI, if at all. Still stress-testing locally to see whether it fixes the crash...

[AlexWaygood]

I let my repro script from #11220 (comment) run for over two hours with this PR branch (which on my slow laptop was enough for 26 iterations). There were no crashes. Prior to #11343, like @hauntsaninja reported in #11343 (comment), I was seeing frequent crashes involving resource.struct_rusage, and less frequent crashes involving other classes. I would always have seen a crash before 26 iterations.

If somebody else wants to do more stress-testing, feel free, but I need to reclaim my laptop for other tasks now, since stress-testing this PR uses all of my laptop's cores 😄

[JelleZijlstra]

Incremental mode probably doesn't help in CI anyway since there won't be any cache.

[AlexWaygood]

Incremental mode probably doesn't help in CI anyway since there won't be any cache.

Yeah, and performance doesn't matter nearly as much for local usage, since I doubt anybody's running the script with --all locally (I certainly am not)