[llvm] [AMDGPU] Use table strategy for LowerModuleLDSPass at O0 (PR #160181)

[Jon Chesterfield via llvm-commits]

JonChesterfield wrote:

The different strategies change how variables are accessed but not where they are allocated. This change may improve compile time. It will definitely regress runtime.

If it changes the the reported amount of lds used by any kernel, there is an error in this pass or elsewhere.

https://github.com/llvm/llvm-project/pull/160181