Federated Prompting and Chain-of-Thought Reasoning for Improving LLMs Answering