feat(examples): add TableBot LangGraph FCP endpoint

[chaliy]

What

Add a uv-based examples/langgraph project that implements a concrete FCP target actor named TableBot. The example uses FastAPI, LangGraph memory, and OpenAI through langchain-openai.

Why

The protocol benefits from an example where a client actor can discover capabilities, negotiate authentication, and complete a multiturn task without knowing an API schema upfront.

How

• Add a public GET / handshake describing TableBot capabilities and the demo auth scheme.
• Add POST / text/JSON handling with fcp_session cookie-backed LangGraph memory.
• Pass per-request demo auth context from Authorization: Demo <your name> into the model call.
• Include tests for handshake content, session reuse, auth context, body parsing, and model failure handling.

Risk

• Low
• Example-only change; runtime behavior depends on a valid OPENAI_API_KEY and model quota.

Checklist

• Tests added or updated
• Backward compatibility considered

Validation: uv run pytest from examples/langgraph.