The stochastic and dynamic inventory routing problem with fine time granularity: A reinforcement learning approach