List Items One by One: A New Data Source and Learning Paradigm for Multimodal LLMs