Add intrinsics for sin, cos, and pow. These use llvm_anyfloat_ty, and so