[llvm] [GVN] Permit load PRE to happen in more cases (PR #76063)

[David Sherwood via llvm-commits]

david-arm wrote:

> You should start by trying to understand why GVN does not recognize the available load in that test case. It probably has something to do with the the fact that one block contains two available loads for different phi translated pointers.

OK thanks. I think I see now that `PredLoads` is different to what I thought. If I understand correctly, `IsValueFullyAvailableInBlock` should return true for my test, which would then prevent `PredLoads` from being populated.

https://github.com/llvm/llvm-project/pull/76063