[llvm] AMDGPU/GlobalISel: Fold (trunc (lshr x, 16)) to unmerge high half (PR #217633)

[Joe Nash via llvm-commits]

https://github.com/Sisyph commented:

I'm not very knowledgeable of those implementation details, but the test changes look quite good. LGTM

https://github.com/llvm/llvm-project/pull/217633