[llvm] [AMDGPU][NFC] Refactor D16 folding for image samples with multiple ExtractElement+FPTrunc chains (PR #145312)

[Harrison Hao via llvm-commits]

harrisonGPU wrote:

Because I want to support D16 folding for image instructions involving multiple extractelement and pkrtz operations, I need to refactor this function. After the QA team finishes testing, I will push a PR to enable D16 folding support for such cases.

https://github.com/llvm/llvm-project/pull/145312