TEA: Trajectory Encoding Augmentation for Robust and Transferable Policies in Offline Reinforcement Learning [2411.19133]