Send serialized ASTs to parallel workers

[ilevkivskyi]

This way we can properly benefit from native parser in parallel type checking. Self-check is now ~2.2x faster with 5 workers compared to in-process checking (also with native parser). Also it uses less memory, but still with 5 workers, self-check takes ~twice more memory compared to in-process.

Implementation is mostly straightforward. The GC freeze hack needed some tuning, as there is no single hot-spot in terms of allocations anymore.

Note: do not use maturin develop for any performance measurements, as this creates some very slow wheel.

[ilevkivskyi]

Ping on this one. I want to continue with other improvements to parallel checking that depends on this PR.

[JukkaL]

Sorry, was busy today, but I should be able to test this with our big internal codebase tomorrow. Overall looks good, I just want to try this out a little bit.

[github-actions]

According to mypy_primer, this change doesn't affect type check results on a corpus of open source code. ✅

[JukkaL]

I didn't have time to try this out with the internal repo, but I can do it after this has been merged. Approving to unblock follow-up work. Anyway, great results so far -- I think this already makes parallel type checking quite useful, even if we can continue to improve scalability further!