Support for Prompt Caching in Completion Usage

[Congyuwang]

https://platform.openai.com/docs/guides/prompt-caching

"usage": {
  "prompt_tokens": 2006,
  "completion_tokens": 300,
  "total_tokens": 2306,
  "prompt_tokens_details": {
    "cached_tokens": 1920
  },
  "completion_tokens_details": {
    "reasoning_tokens": 0
  }
}