Before this commit `-C codegen-units` would just get silently be
ignored if `-C incremental` was specified too. After this commit
one can control the number of codegen units generated during
incremental compilation. The default is rather high at 256, so most
crates won't see a difference unless explicitly opting into a lower
count.

…:new.

…atsakis

Make the rustc respect the `-C codegen-units` flag in incremental mode.

This PR implements (the as of yet unapproved) major change proposal at rust-lang/compiler-team#245. See the description there for background and rationale.

The changes are pretty straightforward and should be easy to rebase if the proposal gets accepted at some point.

r? @nikomatsakis cc @pnkfelix

Implement Hash for Infallible

https://www.reddit.com/r/rust/comments/fmllgx/never_crate_stable_alternative_to/ lists not implementing `Hash` as a reason for the `never` crate. I see no reason not to implement `Hash` for `Infallible`, so might as well do it.

No changes necessary for `!`, because `!` already implements `Hash` (see rust-lang#51404).

…fix, r=Amanieu

Fix double-free and undefined behaviour in libstd::syn::unix::Thread::new

While working on concurrency support for Miri, I found that the `libstd::syn::unix::Thread::new` method has two potential problems: double-free and undefined behaviour.

**Double-free** could occur if the following events happened (credit for pointing this out goes to @RalfJung):

1.  The call to `pthread_create` successfully launched a new thread that executed to completion and deallocated `p`.
2.  The call to `pthread_attr_destroy` returned a non-zero value causing the `assert_eq!` to panic.
3.  Since `mem::forget(p)` was not yet executed, the destructor of `p` would be executed and cause a double-free.

As far as I understand, this code also violates the stacked-borrows aliasing rules and thus would result in **undefined behaviour** if these rules were adopted.  The problem is that the ownership of `p` is passed to the newly created thread before the call to `mem::forget`. Since the call to `mem::forget` is still a call, it counts as a use of `p` and triggers UB.

This pull request changes the code to use `mem::ManuallyDrop` instead of `mem::forget`. As a consequence, in case of a panic, `p` would be potentially leaked, which while undesirable is probably better than double-free or undefined behaviour.