[clang] [LifetimeSafety] Count escape facts when classifying persistent origins (PR #204485)

[Gábor Horváth via cfe-commits]

Xazax-hun wrote:

> Why do we not do this for InvalidateOrigin, Expire and other fact kinds, can origins associated with them not be persistent?

Good question. I think the answer might be that we also need a use or escape for these facts to trigger an error, so that use/escape would make those origins persistent. At least, I did not see a false negative that can be root caused to those origins not marked as such just yet. 

https://github.com/llvm/llvm-project/pull/204485