[llvm] [AMDGPU][CodeGen] Do not rematerialize registers with convergent users (PR #222322)

[Lucas Ramirez via llvm-commits]

lucas-rami wrote:

> Is this related to https://github.com/llvm/llvm-project/issues/222368?

This looks like the same sort of issue conceptually (rematerialization to an invalid position) but this patch won't resolve it unfortunately; this deals with rematerialization in the scheduler not the register allocator. 

https://github.com/llvm/llvm-project/pull/222322