Do LLMs Know What Other LLMs Don't? Peer-Probing as Scalable Evaluation