Change the generation of the vmuluwm instruction to be based on the MUL opcode.
The vmuluwm instruction is used to multiply two v4i32 vectors and produce a v4i32 vector result. The original implementation used an intrinsic to do this, however it can be done using the multiply opcode in LLVM.
This requires the definition of the vmuluwm to be changed from an intrinsic to a VXForm_1 instruction. It also requires the MUL instruction lowering for the v4i32 types to be changed from Custom to Legal when the POWER8 Altivec feature is available.