Open-World Reinforcement Learning over Long Short-Term Imagination