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

[Jameson Nash via llvm-commits]

vtjnash wrote:

We need to stop blocking optimization when everything is correct except for an extra lifetime marker, since it is not correct to block optimization just because some information is incomplete, and I have already prepared a follow up commit where size ceases to be available and optimizes many more cases. I just didn't want to combine the two changes if possible

As for restoring lifetimes, I agree that is desirable, and I don't think this has any impact on it, since failing to execute the optimization when applicable and failing to optimize the lifetime markers after the optimization don't seem like they should block each other.

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