Reinforcement Learning of Spatio-Temporal Point Processes