feat: add GPT-OSS 120b and 20b models to Groq provider

[daniel-lxs]

Add OpenAI's GPT-OSS models to Groq provider:

• GPT-OSS 120b: MoE architecture with 128 experts, 131K context
• GPT-OSS 20b: MoE architecture with 32 experts, 131K context

Both models support tool use, browser search, code execution, and JSON object mode.

---

Important

Add GPT-OSS models to Groq provider and update reasoning budget logic for Gemini 2.5 Pro models.

• Models:
  • Add openai/gpt-oss-120b and openai/gpt-oss-20b to groq.ts with specific parameters.
• Reasoning Budget:
  • Update getModelParams in model-params.ts to handle Gemini 2.5 Pro models with a minimum thinking budget of 128 tokens.
  • Adjust ThinkingBudget.tsx to use a minimum of 128 tokens for Gemini 2.5 Pro models.
• Tests:
  • Add tests in model-params.spec.ts and ThinkingBudget.spec.tsx to verify the new reasoning budget logic for Gemini 2.5 Pro models.

^{This description was created by for 4480cb5. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Thank you for your contribution! I've reviewed the changes and found several issues that need attention before merging.

[roomote]

The PR description mentions that these models support tool use, browser search, code execution, and JSON object mode. Should we add these as boolean flags in the model info? For example:

[roomote]

Is this description correct? The model name suggests 20B parameters total, but the description says "20 billion parameters and 32 experts". Should this be "GPT-OSS 20B is OpenAI's flagship open source model, built on a Mixture-of-Experts (MoE) architecture with 32 experts"?

[daniel-lxs]

yes, shut up

[roomote]

Could you add a comment explaining why 128 tokens was chosen as the minimum for Gemini 2.5 Pro models? This is a significant reduction from the standard 1024 minimum and might affect model performance.

[roomote]

This change from 1024 to 128 minimum thinking tokens for Gemini 2.5 Pro models could be a breaking change for existing users. Is this intentional? Also, the PR description doesn't mention these Gemini changes - should they be in a separate PR?