Support other finish reasons for OpenAiClient (deepseek)

[thomasht86]

I confirm that this contribution is made under the terms of the license found in the root directory of this repository's source tree and that I have the authority necessary to make this contribution on behalf of its copyright owner.

When hitting OpenAI-compatible endpoints, such as https://docs.together.ai/docs/openai-api-compatibility, other models may often have other finish_reasons than OpenAI's models, based on their stop-tokens.

Examples:

• meta-llama/Llama-3.3-70B-Instruct-Turbo:

curl -X POST "https://api.together.xyz/v1/chat/completions" \
  -H "Authorization: Bearer $TOGETHER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "meta-llama/Llama-3.3-70B-Instruct-Turbo",
    "messages": [],
    "max_tokens": null,
    "temperature": 0.7,
    "top_p": 0.7,
    "top_k": 50,
    "repetition_penalty": 1,
    "stop": ["<|eot_id|>","<|eom_id|>"],
    "stream": true
  }'

• deepseek-ai/DeepSeek-R1:

curl -X POST "https://api.together.xyz/v1/chat/completions" \
 -H "Authorization: Bearer $TOGETHER_API_KEY" \
 -H "Content-Type: application/json" \
 -d '{
   "model": "deepseek-ai/DeepSeek-R1",
   "messages": [],
   "max_tokens": null,
   "temperature": 0.6,
   "top_p": 0.95,
   "top_k": 50,
   "repetition_penalty": 1,
   "stop": ["<｜end▁of▁sentence｜>"],
   "stream": true
 }'

Before this PR, a request to this endpoint would result in:

"fields": {
    "error": "Unknown OpenAi completion finish reason 'eos'",
    "source": "openai",
    "code": 400
}

This PR changes the default behavior to only throw exception on finish_reason=error, and let through other ones, making our OpenAiClient more generic.

[glebashnik]

Looks good!