Deep Reinforcement Learning In The Imperfect-Information Game Coup