Text this: Optimization of fluid control laws through deep reinforcement learning using dynamic mode decomposition as the environment