[llvm] [Coro] Amortize debug info processing cost in CoroSplit (PR #109032)

[Artem Pianykh via llvm-commits]

artempyanykh wrote:

This has now been fully merged. I updated the description with references to relevant PRs. The final implementation turned out faster than the initial one: in terms of the CoroSplit pass runtime on a sample cpp file with `-g2`: 306ms -> 3.8ms @ 80x now vs 306ms -> 17ms @ 18x which is almost as fast as `-g1`.

Thanks @felipepiovezan, @TylerNowicki, and @adrian-prantl for reviews!

https://github.com/llvm/llvm-project/pull/109032