feat(runtimes): support dynamic model pricing from OpenRouter and env overrides

[RainSunMe]

What

Extend the runtime cost estimation to support all LLM providers, not just Anthropic Claude.

Why

Currently, MODEL_PRICING in packages/views/runtimes/utils.ts only contains hardcoded Claude model prices. Users running non-Anthropic models (MiMo, DeepSeek, GPT, etc.) see "Tokens recorded but pricing missing" with no way to resolve it without submitting a PR upstream.

How

Three pricing sources are merged at runtime, in descending priority:

1. Hardcoded pricing — existing Claude entries remain unchanged as the authoritative source for Anthropic models.
2. Environment variable — NEXT_PUBLIC_MODEL_PRICING_OVERRIDES accepts a JSON object for custom or corrected pricing:

   {"custom/model":{"input":2,"output":6,"cacheRead":0.5,"cacheWrite":0}}

3. OpenRouter API — on page load, pricing for 500+ models is fetched from https://openrouter.ai/api/v1/models and merged for any model not already covered by the above.

Only models missing from the first two layers are filled in from OpenRouter, so hardcoded and user-specified prices always take precedence.

Testing

• Verified that existing Claude pricing is unaffected.
• Confirmed MiMo and other non-Claude models resolve to correct costs after OpenRouter fetch.
• NEXT_PUBLIC_MODEL_PRICING_OVERRIDES correctly overrides both hardcoded and OpenRouter entries.

[vercel]

@RainSunMe is attempting to deploy a commit to the IndexLabs Team on Vercel.

A member of the Team first needs to authorize it.