Selective Replay Enhances Learning in Online Continual Analogical Reasoning