Default models updated across several providers

Cerebras and CerebrasOpenAI now default to gpt-oss-120b (previously llama-4-scout-17b-16e-instruct), Gemini defaults to 3.7 Flash, and Groq's deprecated llama-3.3-70b-versatile default is now openai/gpt-oss-120b. If you set a model id explicitly, nothing changes for you; only the defaults moved.