[llvm] [AArch64][ISel] Enable profile-aware branch condition merging (PR #201486)

[Kunal Pathak via llvm-commits]

kunalspathak wrote:

> It looks like this regresses 709.cactus_r by ~8% on Neoverse V2 with `-O3`/`-O3 -mcpu=neoverse-v2`. Other options show similar slowdowns, except with `-ffast-math`. Have you seen anything similar?

I didn't see it. Let me take a look.

https://github.com/llvm/llvm-project/pull/201486