docs(examples): LangChain agent stopped at its budget

[prashar32]

What

A runnable LangChain example — examples/langchain/. Wrap a LangChain LLM loop with RiskKernelCallbackHandler and the deterministic governor caps it: one governed step per model call, hard-stopped at the loop budget, with the halt propagating out of llm.invoke() to end the chain.

🛑 RiskKernel halted the LangChain run — reason: loop_budget_exceeded
   ── final ledger (enforced by the governor) ──
     LLM calls (loops) :    6   (budget: 6)

Key-free, like wrap-your-agent

Uses LangChain's FakeListLLM, so the loop enforcement runs with nothing but riskkernel serve — no model call, no spend. The README shows the few lines to route a real ChatOpenAI/ChatAnthropic through run.proxy_config() for the dollar/token ceiling, and points to codebase-qa for the real-model dollar kill.

Stacked on #55

This example only actually stops because of the raise_error=True fix in #55 — without it LangChain swallows the halt and the loop runs on. Base is set to the #55 branch; merge #55 first and this will retarget to main (diff becomes example-only).

Verification

Ran end-to-end against a keyless daemon with langchain-core 1.4.0: 6 calls allowed, the 7th halted at loop_budget_exceeded (exit 0). Error paths checked too — daemon-down prints a start-it hint (exit 1); missing langchain-core prints an install hint (exit 1). py_compile clean.