[ARM][MVE] Fix for tail predication for loops containing MVE gather/scatters

[anwel]

Summary

Fix to include non-predicated version of write-back gather in special case treatment for deducting the instruction type.
(This is fixing https://reviews.llvm.org/D85138 for corner cases)

[anwel]

Added a test that would fail the VecTy && "No scalable vectors expected here" assert without this fix.

[dmgreen]

OK. Thanks. LGTM

llvm/lib/Target/ARM/MVETailPredication.cpp

 
   return true;
 }
 
 static FixedVectorType *getVectorType(IntrinsicInst *I) {
   unsigned ID = I->getIntrinsicID();
   FixedVectorType *VecTy;
   if (ID == Intrinsic::masked_load || isGather(I)) {
-    if (ID == Intrinsic::arm_mve_vldr_gather_base_wb_predicated)
+    if (ID == Intrinsic::arm_mve_vldr_gather_base_wb ||
+        ID == Intrinsic::arm_mve_vldr_gather_base_wb_predicated)
       // then the type is a StructType
       VecTy = dyn_cast<FixedVectorType>(I->getType()->getContainedType(0));
     else
       VecTy = dyn_cast<FixedVectorType>(I->getType());
   } else if (ID == Intrinsic::masked_store) {
     VecTy = dyn_cast<FixedVectorType>(I->getOperand(0)->getType());
   } else {
     VecTy = dyn_cast<FixedVectorType>(I->getOperand(2)->getType());