Reinforcement learning and Bayesian data assimilation for model-informed precision dosing in oncology