For ".vf" instructions, there is a floating point scalar argument for the operations. If the scalar type of the argument is f64, we need to assign FPR64 for it.
In this patch, I add "CodeGenOnly" instructions for the FPR64 version of the instructions. For example, there are two instructions for vfadd.vf. One is VFADD_VF for FPR32 and the other is VFADD_VD for FPR64.