[llvm] [MemCpyOpt] Handle scalable aggregate types in memmove/memset formation (PR #80487)

[Philip Reames via llvm-commits]

preames wrote:

> @preames I believe this is a candidate for backporting? (the test case does indeed seem to crash on the release/18.x branch right now).

It could be, but I don't know we need to.  It was a fuzzer failure, nothing from a real workload.  

https://github.com/llvm/llvm-project/pull/80487