[clang] [llvm] [VectorCombine] Trim low end of loads used in shufflevector rebroadcasts. (PR #149093)

[Yingwei Zheng via cfe-commits]

dtcxzyw wrote:

FYI this patch may cause some failures on llvm-opt-benchmark: https://github.com/dtcxzyw/llvm-opt-benchmark/issues/3405. I will provide a reproducer later.

https://github.com/llvm/llvm-project/pull/149093