Policy Poisoning in Batch Reinforcement Learning and Control