Generate then Retrieve: Conversational Response Retrieval Using LLMs as Answer and Query Generators