CL: Fix vstore_local on Intel
Doing a RMW on a vector is not atomic, and that's exactly what vtn_local_store() does when it sees a vector that's cast to an array on which one entry is then updated. Let's keep the store as-is in that case.
Doing a RMW on a vector is not atomic, and that's exactly what vtn_local_store() does when it sees a vector that's cast to an array on which one entry is then updated. Let's keep the store as-is in that case.