[llvm] [AMDGPU][SIInsertWaitcnts] Do not add s_waitcnt when the counters are known to be 0 already (PR #72830)

[Carl Ritson via llvm-commits]

perlfu wrote:

@Pierre-vh I am seeing some regressions from this change causing generation of extra unnecessary `s_waitcnt` instructions.

See attached test: `llc -march=amdgcn -start-before=si-insert-waitcnts -mcpu=gfx1030 -verify-machineinstrs -mattr=+wavefrontsize32,-wavefrontsize64 -o - test.mir`

```
	s_waitcnt vmcnt(0)
	v_mov_b32_e32 v0, 0
	s_waitcnt vmcnt(1)                   <-- new extra s_waitcnt
	v_readfirstlane_b32 s8, v7
	v_readfirstlane_b32 s9, v8
	v_readfirstlane_b32 s10, v9
	v_readfirstlane_b32 s11, v10
	v_readfirstlane_b32 s12, v11
	v_readfirstlane_b32 s13, v12
	v_readfirstlane_b32 s14, v13
	v_readfirstlane_b32 s15, v14
	s_waitcnt vmcnt(0)                   <-- new extra s_waitcnt
	v_readfirstlane_b32 s16, v15
	v_readfirstlane_b32 s17, v16
	v_readfirstlane_b32 s18, v17
	v_readfirstlane_b32 s19, v18
	v_mov_b32_e32 v1, v0
```

[test.mir.gz](https://github.com/llvm/llvm-project/files/13695014/test.mir.gz)

https://github.com/llvm/llvm-project/pull/72830