[llvm] [AMDGPU][SIInsertWaitCnts] Use RegUnits-based tracking (PR #162077)

[Pierre van Houtryve via llvm-commits]

Pierre-vh wrote:

> Maybe this should use SparseBitVector?

It stores counters, not single bits. I think using a `SparseBitVector` will just make this more complicated than it has to be
Is there a good reason why DenseMap isn't good here? 

https://github.com/llvm/llvm-project/pull/162077