Explaining User Performance in Information Retrieval: Challenges to IR Evaluation