Fix Requesty extended thinking

[cte]

Description

The getModelParams function is the source of truth for generating reasoning effort and extended thinking parameters to pass to the model.

---

Important

Centralizes model parameter logic in getModelParams for RequestyHandler, MistralHandler, and OpenAiHandler, updating tests to ensure correct parameter usage.

• Behavior:
  • Centralizes model parameter logic in getModelParams for RequestyHandler, MistralHandler, and OpenAiHandler.
  • Updates getModel() in requesty.ts, mistral.ts, and openai.ts to use getModelParams for maxTokens and temperature.
  • Ensures RequestyHandler includes reasoning_effort and thinking parameters from getModelParams.
• Tests:
  • Updates requesty.test.ts to verify correct max_tokens and temperature in createMessage and completePrompt.
  • Modifies deepseek.test.ts to ensure DeepSeekHandler initializes with correct API key and model parameters.
• Misc:
  • Removes redundant logic for setting maxTokens and temperature in mistral.ts and openai.ts.

^{This description was created by for 344b2e8. You can customize this summary. It will automatically update as commits are pushed.}

[cte]

CC @dtrugman

[cte]

It looks like Requesty is using the Anthropic format for extending thinking.

[cte]

It seems we were defaulting to 0 for any provider that made use of getModelParams (including OpenRouter). I'm not 100% sure what the implications of this are for Requesty.