[llvm-branch-commits] [IR] Make BranchInst operand order consistent (PR #151673)

[Alexis Engelke via llvm-branch-commits]

aengelke wrote:

This causes more breakage (would need to investigate, but there're >100 test failures for not all-the-same reasons). But what would change? `Op<-1>` is `op_end() - 1`, `Op<0>` is `op_begin()` which is `op_end() - getNumOperands()` which is essentially the same as what this PR currently has. getNumOperands is expensive due to the load+bitfield extract.

https://github.com/llvm/llvm-project/pull/151673