[llvm] [AMDGPU] Use a range-based for loop (NFC) (PR #148767)

[Matt Arsenault via llvm-commits]

================
@@ -3981,8 +3981,8 @@ bool AMDGPUAsmParser::validateVOPD(const MCInst &Inst,
   bool AsVOPD3 = MII.get(Opcode).TSFlags & SIInstrFlags::VOPD3;
 
   if (AsVOPD3) {
-    for (unsigned I = 0, E = Operands.size(); I != E; ++I) {
-      AMDGPUOperand &Op = ((AMDGPUOperand &)*Operands[I]);
+    for (const auto &Operand : Operands) {
+      AMDGPUOperand &Op = (AMDGPUOperand &)*Operand;
----------------
arsenm wrote:

No auto, it's extra confusing if we need a cast right after 

https://github.com/llvm/llvm-project/pull/148767