Opportunities for reinforcement learning in stochastic dynamic vehicle routing