Remove InitializeNonlocalInstruction

[rdmarsh2]

This removes InitializeNonlocalInstruction from the IR, which should result in a small performance improvement in building the IR. I think the consistency checks need to be adjusted to handle phi nodes for global and static variables which don't have their address taken, but I'm not sure exactly how to identify those.

[jbj]

It's great to see that this diff takes away more than it adds.

[MathiasVP]

Yay for fewer instructions 🎉 ! Apparently, GitHub doesn't allow comments on collapsed lines, so I'm leaving this as a comment instead:

I guess now that we no longer have an InitializeNonLocal instruction this comment on the canDefineReadOnly override in AllNonLocalMemory doesn't make sense anymore, right?

  override predicate canDefineReadOnly() {
    // A "must" access that defines all non-local memory appears only on the `InitializeNonLocal`
    // instruction, which provides the initial definition for all memory outside of the current
    // function's stack frame. This memory includes string literals and other read-only globals, so
    // we allow such an access to be the definition for a use of a read-only location.
    not isMayAccess()
  }

[rdmarsh2]

Yeah, I think I can just make that override predicate canDefineReadOnly() { none() } now.

[jbj]

The tests are passing. Is it time to take this out of draft mode?

[dbartol]

LGTM (@rdmarsh2 and I had already discussed most of these changes via Zoom).

[dbartol]

LGTM