Tell Me Where You Are: Multimodal LLMs Meet Place Recognition