Learning Vision-based Robotic Manipulation Tasks Sequentially in Offline Reinforcement Learning Settings