AArch4: Fix range loop contradicting comment above it

[john.brawn]

It's possible that the blocks that are created end up never needing branch relaxation in which case not updating end() doesn't matter, but it's not clear from a quick look if that's true or not and just doing what the comment says seems perfectly reasonable. LGTM.

[arsenm]

r272959

lib/Target/AArch64/AArch64BranchRelaxation.cpp

bool AArch64BranchRelaxation::fixupConditionalBranch(MachineInstr *MI) {
   adjustBlockOffsets(*MBB);
   return true;
 }
 
 bool AArch64BranchRelaxation::relaxBranchInstructions() {
   bool Changed = false;
   // Relaxing branches involves creating new basic blocks, so re-eval
   // end() for termination.
-  for (auto &MBB : *MF) {
+  for (MachineFunction::iterator I = MF->begin(); I != MF->end(); ++I) {
+    MachineBasicBlock &MBB = *I;
     MachineInstr *MI = MBB.getFirstTerminator();
     if (isConditionalBranch(MI->getOpcode()) &&
         !isBlockInRange(MI, getDestBlock(MI),
                         getBranchDisplacementBits(MI->getOpcode()))) {
       fixupConditionalBranch(MI);
       ++NumRelaxed;
       Changed = true;
     }