[RISCV] Handle scalarized reductions in getArithmeticReductionCost

This fixes a crash reported at
llvm#114250 (comment)

If the vector type isn't legal at all, e.g. bfloat with +zvfbfmin,
then the legalized type will be scalarized. So use getScalarType()
instead of getVectorElement() when checking for f16/bf16.

(cherry picked from commit 053451c)

Author: lukel97

llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp

@@ -1658,9 +1658,8 @@ RISCVTTIImpl::getArithmeticReductionCost(unsigned Opcode, VectorType *Ty,
     break;
   case ISD::FADD:
     // We can't promote f16/bf16 fadd reductions.
-    if ((LT.second.getVectorElementType() == MVT::f16 &&
-         !ST->hasVInstructionsF16()) ||
-        LT.second.getVectorElementType() == MVT::bf16)
+    if ((LT.second.getScalarType() == MVT::f16 && !ST->hasVInstructionsF16()) ||
+        LT.second.getScalarType() == MVT::bf16)
       return BaseT::getArithmeticReductionCost(Opcode, Ty, FMF, CostKind);
     if (TTI::requiresOrderedReduction(FMF)) {
       Opcodes.push_back(RISCV::VFMV_S_F);