feat: Configurable model usage with runtime AI provider selection

[Copilot]

Adds runtime AI provider configuration, allowing users to override environment defaults with custom API keys and model selections through the settings UI. Configuration is memory-only (not persisted) and supports OpenAI, Google Gemini, and Anthropic.

Changes

Frontend

• Settings context manages AI vendor, API key, and model name state
• Settings modal UI with real-time API key validation
• Validation feedback for incompatible vendor/model combinations

Backend

• /api/validate-key endpoint validates credentials across providers
• All LiteLLM agent calls accept optional custom_api_key and custom_model parameters
• Pydantic models updated: ChatRequest, ConfirmRequest include optional custom fields

Configuration

• AI_VENDORS mapping defines supported providers and example models
• Custom settings passed directly to LiteLLM via extra_body parameter
• Falls back to environment configuration when custom settings absent

# Example agent usage with custom configuration
response = await analysis_agent(
    question="SELECT * FROM users",
    graph_schema=schema,
    custom_api_key="sk-...",
    custom_model="gpt-4"
)

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.