The builtin for vec_orc has support for the following two signatures,
but currently the compiler marks it ambiguous:
vector float vec_orc(vector float, vector float)
vector double vec_orc(vector double, vector double)
This patch implements these two builtins.
nit: line too long?