Add LiteLLM support

[ocs-percy]

Add LiteLLM Support

Integrate LiteLLM as a supported LLM provider/proxy in Open Chat Studio.

LiteLLM provides a unified interface to 100+ LLMs and can be self-hosted as a proxy, making it popular for teams that want to centralise model access, add spend controls, and maintain a consistent API surface regardless of underlying provider.

What needs to happen

• Add LiteLLM (proxy) as a provider option in the model configuration UI
• Support configuring a custom LiteLLM base URL and API key per team/org
• Map available models from the LiteLLM /models endpoint (or allow manual entry)
• Handle LiteLLM-specific behaviours (e.g. streaming differences, error formats)
• Test against a self-hosted LiteLLM instance with multiple backend models

Notes

LiteLLM exposes an OpenAI-compatible API, so this may share a code path with the OpenRouter integration. Consider whether both can be handled under a generic "OpenAI-compatible endpoint" provider type.