[llvm] [AMDGPU] Handle bf16 operands the same way as f16. NFC. (PR #77826)

[Matt Arsenault via llvm-commits]

================
@@ -284,12 +284,17 @@ def SIfptrunc_round_downward : SDNode<"AMDGPUISD::FPTRUNC_ROUND_DOWNWARD",
 // Returns 1 if the source arguments have modifiers, 0 if they do not.
 class isFloatType<ValueType SrcVT> {
   bit ret = !or(!eq(SrcVT.Value, f16.Value),
----------------
arsenm wrote:

ValueType has an isFP field; we could probably get rid of isFloatType altogether 

https://github.com/llvm/llvm-project/pull/77826