On the rank of cyclic latin squares