[llvm] [InstCombine] Push freeze through non-recurrence PHIs (PR #157678)

[Nikita Popov via llvm-commits]

nikic wrote:

> I also collected data on the # of freezes in SPEC2017 (Ofast, LTO) after this patch and there's a ~13% increase. #154336 also caused an increase of ~27%. Although as previously mentioned [#154336 (comment)](https://github.com/llvm/llvm-project/pull/154336#issuecomment-3201408296), I'm not convinced this is a particularly useful metric, but just wanted to share this.

The tradeoff here is basically between having less freeze or having more freeze in positions where it has less impact. That the total number of freezes increases is expected.

https://github.com/llvm/llvm-project/pull/157678