[llvm] [MemCpyOpt] fix incorrect handling of lifetime markers (PR #143782)

[Nikita Popov via llvm-commits]

nikic wrote:

Just like the previous changes, this still only improves handling for the case where the lifetime size does not match the alloca size, which should not occur in practice. Why do we need to handle it?

Note that this code originally preserved lifetime markers (roughly based on the NCD of the lifetimes) and that got "temporarily" dropped due to some issues and never restored. I don't think the current code cares about the size, but if we wanted to restore this (which we really should...) the size would matter again.

https://github.com/llvm/llvm-project/pull/143782