Initial cmov crate

[tarcieri]

Implements safe wrappers for the cmovz and cmovnz instructions which work on x86 and x86_64, implemented using inline assembly.

Using inline assembly allows us to provide a sort of black box which LLVM will not interfere with, which is otherwise problematic when using anything besides ASM.

It's otherwise not possible to correctly emit CMOV instructions on x86 platforms with LLVM because the x86-cmov-conversion pass which will potentially rewrite them as branches.

For more details, see:

https://dsprenkels.com/cmov-conversion.html

[tarcieri]

Went ahead and added a portable fallback implementation for non-x86/x86_64 platforms in 5dfbdca with optimizations in 876653d which uses bitwise arithmetic but can't otherwise specifically guarantee constant-time operation.

I think it's nice to have from an ergonomics perspective, even if we can't make guarantees.

We could potentially add some tests that automatically inspect the generated assembly and make sure the generated instructions fall within a particular allow list. That'd provide a degree of due diligence that the resulting implementation is constant time, if not a guarantee.