feat(cookbooks): add live LangGraph e2e recipe through openinference adapter - #4

What

Adds cookbooks/live_langgraph.py: a live, end-to-end LangGraph example. It builds a real ReAct agent (create_react_agent + gpt-4o-mini), instruments it with openinference-instrumentation-langchain, exports genuine OTel spans in-memory, and diffs two real graph executions through the existing openinference adapter — no manual trace reshaping.

Why

Roadmap item J3 — context/ROADMAP.md WS-J: "Real end-to-end examples (LangGraph / CrewAI traces)". LangGraph is the most-adopted agent framework; this proves AgentDiff ingests its telemetry natively and gates regressions on real graph output (verified live).

How

• Baseline run: single efficient DB lookup. Candidate run: same task across NY/CA/TX → extra tool iterations.
• Spans sorted chronologically before projection; AGENT/CHAIN → routing, LLM → llm_call, TOOL → tool_call.
• Live verification (2026-08-22): candidate showed TDI 0.118, 2 loop flags (node-level + tool-level), token delta +99.54%, gate correctly BLOCKED.
• Also updates cookbooks/README.md table + run commands and adds a CHANGELOG [Unreleased] entry.

Testing

• Ran the recipe live against the real API: gate blocked as designed (output recorded in context/LIVE-VERIFICATION.md)
• make lint -> pass; uv run python -m pytest -> 205 passed

Checklist

• Exactly one logical change in this PR
• make lint passes
• uv run python -m pytest is green (205 passed)
• Docs updated if user-facing
• CHANGELOG.md entry added under [Unreleased]
• Conventional title: feat(cookbooks): add live LangGraph e2e recipe through openinference adapter