fix: fix format_response for better API response handling

[CheinTian]

fix format_response method to improve handling of API responses, including embedding and reasoning content.

[CheinTian]

makes format more robust. when reasoning_content is null, do not add <think></think> label.

Here is a example:

{ "index": 0, "message": { "role": "assistant", "content": "你好，我是DeepSeek，一个由深度求索公司创造的AI助手，很高兴为你提供帮助！😊", "refusal": null, "annotations": null, "audio": null, "function_call": null, "tool_calls": [], "reasoning_content": null }, "logprobs": null, "finish_reason": "stop", "stop_reason": null, "token_ids": null }

old version add label when reasoning_content is null. we don't want it.

[scuuy]

Thanks for this code part😊, it indeed becomes more robust without changing the former usage. We will merge it.