feat: support gpt-4o-audio

[Laisky]

功能

支持 gpt-4o-audio 系列的 chat completetions

暂时没搞懂 stream 的计费模式，所以只支持了 stream==false

合并前可以使用 https://github.com/Laisky/one-api 来体验这个功能。

依赖

本提交也包含了 #2022

自测

[WqyJh]

对于 stream 请求，请求体中加入 include_usage 参数，返回的其中一条消息会包含 usage。建议

• fastpath: 直接修改请求体，强制插入这个参数，然后记录返回的 usage。
• slowpath: 自己算

"stream_options": {
  "include_usage": true
},

data: {"id":"chatcmpl-AqcNUZRReOSzmK4Bqntp7s5KwzaMR","object":"chat.completion.chunk","created":1737103944,"model":"gpt-4o-audio-preview-2024-12-17","service_tier":"default","system_fingerprint":"fp_58887f9c5a","choices":[],"usage":{"prompt_tokens":17,"completion_tokens":488,"total_tokens":505,"prompt_tokens_details":{"cached_tokens":0,"audio_tokens":0},"completion_tokens_details":{"reasoning_tokens":0,"audio_tokens":409,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}}}