docs(ops): document OrcaRouter as a LiteLLM model provider

[Marc-oss-hub]

Summary

Adds OrcaRouter to the LiteLLM provider docs. OrcaRouter is an OpenAI-compatible model routing gateway that exposes models from OpenAI, Anthropic, Google, DeepSeek, Qwen and others behind a single endpoint and API key. It also runs gateway-level, zero-trust security for AI agents on the same endpoint — screening every prompt/response and governing every tool call on a default-deny basis, with no application code changes.

I'm an engineer on the OrcaRouter team.

What this changes

• docs/src/content/docs/ops/litellm.mdx: adds an ### OrcaRouter subsection under "Supported providers", alongside the existing OpenAI / Azure / Gemini entries, showing the LiteLLMEmbedder wiring with api_base and api_key.
• docs/src/content/docs/ops/entity_resolution.mdx: adds an OrcaRouter row to the LlmPairResolver model strings table, with a one-line note on the namespaced model-id form.

How to use it

from cocoindex.ops.litellm import LiteLLMEmbedder

embedder = LiteLLMEmbedder(
    "openai/openai/text-embedding-3-small",
    api_base="https://api.orcarouter.ai/v1",
    api_key="sk-orca-...",
)

The outer openai/ prefix selects the OpenAI-compatible protocol; the inner openai/… is the router's namespaced model id (passed through verbatim). The same shape works for LlmPairResolver(model="openai/openai/gpt-5.5", ...) in entity resolution.

Verification

• npm run build in docs/ passes (98 pages; the 5 pre-existing missing .md twin notices for index.mdx pages are unchanged by this PR).
• Live connectivity against https://api.orcarouter.ai/v1 with an sk-orca- key:
  • litellm.aembedding("openai/openai/text-embedding-3-small", api_base=..., api_key=...) → 200, 1536-dim vector.
  • litellm.acompletion("openai/openai/gpt-5.5", ...) → 200.
  • Negative control: a bare text-embedding-3-small (no openai/ prefix) → 503 No available channel, which is why the docs show the namespaced form.