[llvm] [AMDGPU] Introduce "amdgpu-sw-lower-lds" pass to lower LDS accesses. (PR #87265)

[via llvm-commits]

b-sumner wrote:

> > We could consider splitting this pass, first lowering of LDS to global memory and then instrumenting newly lowered memory operations as future work.
> 
> This would be much better and we should do this. That was what I wanted at the beginning, and why I wanted this renamed to lower SW LDS. As it is, the name is misleading because it's not software LDS lowering when it's mixing in all the sanitizer handling

But this isn't exactly SW LDS.  We need efficient access to the pointer to the base of simulated LDS and we are storing that in "real" LDS as opposed to adding it to the function ABI.

https://github.com/llvm/llvm-project/pull/87265