native-lib/trace: Intercept mmap and munmap

[nia-e]

Adds support for tracing mmaps in native code and logging that. Notably does not let us call libc::mmap(..) on the Rust side and then munmap it on the native side / the other way; I have ideas for maybe doing that, but they'd be quite a bit more involved since it requires us doing some syscall shimming.

[rustbot]

Thank you for contributing to Miri!
Please remember to not force-push to the PR branch except when you need to rebase due to a conflict or when the reviewer asks you for it.

[RalfJung]

I saw the PR and it's in my queue, but it's a big one so it may take a bit until I have time to dig into this -- sorry.

[nia-e]

Realised this actually has a couple issues that need to be addressed and I could do that in this implementation relatively simply, or rewrite it to be better in every way (so we can call mmap on one side of the ffi boundary and munmap on the other w/ the same pointer, and vice versa) with a tiny tiny bit of inline assembly. I think I'll do that and mark this as a draft

[nia-e]

@rustbot author

[oli-obk]

Nia gave me a crash course in various of those ideas at eurorust. I'll do the first reviews and assign to you (Ralf) once I think it's ready

[rustbot]

☔ The latest upstream changes (possibly e110b81) made this pull request unmergeable. Please resolve the merge conflicts.

[RalfJung]

@nia-e I am going to close this PR since it's become inactive and work has moved to smaller follow-up PRs. Maybe it'd make sense to open an issue summarizing the plan and the current status?