Neighbor Q‐learning based consensus control for discrete‐time multi‐agent systems