[llvm] Enable generic overlapping optimization for memmove (PR #177885)

[Osama Abdelkader via llvm-commits]

osamakader wrote:

sorry for the late reply. I've added volatile memmove tests for the main architectures that avoid unaligned loads.
The MaxStoresPerMemmove limit is still respected. The existing test in `llvm/test/CodeGen/AArch64/max-stores-per-mem-ops.ll` verifies this, it shows that memmove operations exceeding the limit (controlled by `-max-store-memmove` flag) correctly fall back to calling the memmove library function. Our changes don't affect this mechanism, which is enforced in `findOptimalMemOpLowering` before the actual lowering occurs.

https://github.com/llvm/llvm-project/pull/177885