rustup for big refactor; kill most of validation

[RalfJung]

Do not force-push, rustc refers to this commit with rust-lang/rust#53424

Fixes #382
Fixes #439

[RalfJung]

Argh, again new new nightly :(

[RalfJung]

This now fixes the compile-fail tests for memory leaks and mutating statics. :) It also gets rid of some hacks that hook some arbitrary functions even though they have MIR... I will probably pay for this by breaking macOS and/or Windows. Let's see.

[RalfJung]

(Just experimenting with AppVeyor)

[RalfJung]

We finally have a nightly with the refactorings this matches, so I am going to try and make CI pass. Otherwise I think this is ready for review.

@oli-obk do you want be to rebase and maybe squash a little? We ended up backing out the miri updates, so I think none of the commits here is actually ever referenced from rustc.

The only actually functional changes in this PR are:

• Removing the unconditional hook of "std::sys::unix::thread::guard::init", "std::sys::unix::thread::guard::current" and "std::sys::unix::fast_thread_local::register_dtor". Instead, we can now emulate foreign statics so these functions can actually properly execute.
• Hook some more pthread_ functions that initialization needs now, and avoid hooking all pthread:* to avoid papering over unimplemented foreign items.
• Use new mechanism provided by miri engine for handling (mutable) statics.
• Re-use intrinsics and lang item implementations from CTFE where possible.

[oli-obk]

squashing/rebasing seems fine to me, but not necessary

[RalfJung]

Well then I'd say let's get this merged. CI is green. Waiting for your r+ :)

[oli-obk]

lolwat? Are you going to need it again in the future? Or why not just remove it? Too much churn?

[RalfJung]

You mean remove the lifetime? Hm, I didn't remember adding it so I just did not touch it.^^ But also indeed I am not sure yet which extra data I will add again for the Stacked Borrows stuff.