[PAC][libunwind] Investigate potential performance regressions after #173765

_Originally posted by @ojhunt in https://github.com/llvm/llvm-project/pull/173765#pullrequestreview-3619307751_:

I'm also curious about the performance impact of this change - I recall having to fight quite a bit at the time. The test case I was using was an overwhelmingly heavy amount of exception dispatch. The parsing itself should have been overwhelmed by the complexity of the dispatch yet I found significant performance regressions.

I actually wonder if we should make a pointer wrapper that can handle ptrauth types - it would make the behavior much more explicit and would Just Work (tm) as a parameter by making the argument type non-pod, and the inliner can flatten from there.