Add logprobs return in ChatCompletionResponse

[windspirit95]

Hi, this pull request is my small update/modification to add logprobs and top_logprobs arguments when creating chat completion request into llama cpp server (align with OpenAI API template).

Sample request:
POST https://host:port/v1/chat/completions
{
"model": "mixtral-8x7b",
"messages": [
{
"role": "system",
"content": "You are a helpful assistant."
},
{
"role": "user",
"content": "Hello! Could you help me plan for 3-day trip in Hanoi?"
}
],
"logprobs": true,
"top_logprobs": 1,
"max_tokens": 1024,
"temperature": 0.0001,
"stream": false
}

Could you help me to review it? Thanks ^^

[abetlen]

Hey @windspirit95 great contribution, I just changed the default to logprobs=False to match the OpenAI api and added an example to the OpenAPI schema.