dRRT*: Scalable and informed asymptotically-optimal multi-robot motion planning