feat: add LiteLLM provider - #9404

Summary

This PR adds LiteLLM Proxy as a provider. LiteLLM is an open source gateway that you host yourself. It gives one OpenAI-compatible endpoint for many model providers.

With this entry, tools that read this catalog can list LiteLLM by name, with the default local URL.

Changes

• providers/litellm/provider.toml: name, LITELLM_API_KEY, @ai-sdk/openai-compatible, http://localhost:4000/v1 and the docs link.
• providers/litellm/logo.svg: the LiteLLM logo with currentColor and a square viewBox.
• providers/litellm/models/gpt-6-luna.toml: one example alias with base_model = "openai/gpt-6-luna". The reasoning effort values match the OpenAI entry, because LiteLLM forwards reasoning_effort to the backend.

Verification

• bun validate passes. The litellm entry has GPT-6 Luna with the limits from the lab entry.
• bun run test in packages/sdk passes (25 tests).
• bun run build in packages/web passes and copies litellm.svg to dist/logos.

Notes

• The catalog needs at least one model. LiteLLM model names are aliases that each proxy sets in its own config. The example alias works only if the proxy defines it.
• I omit cost because the price depends on the backend.
• bun test in packages/core has 10 failures. They also fail on dev without this change.