[llvm] [VPlan] Replace PhiR operand of ComputeAnyOfResult with VPIRFlags. (PR #175657)

[via llvm-commits]

================
@@ -8762,17 +8762,29 @@ void LoopVectorizationPlanner::addReductionResultComputation(
     VPBuilder::InsertPointGuard Guard(Builder);
----------------
ayalz wrote:

Independent: `addReductionResultComputation()` is already way too long, and this makes it even longer (850 LOC). Would be good to shorten, potentially by canonicalizing AnyOf in-loop chains earlier?

https://github.com/llvm/llvm-project/pull/175657