docs: introduce langchain-runcycles for LangChain agent users

[amavashev]

Summary

Documentation-only update introducing langchain-runcycles (PyPI: langchain-runcycles) as the canonical path for LangChain agent middleware integration with Cycles. Companion to the v0.1.0 release of the new sibling package.

No SDK code changes. No protocol changes.

Files changed

• README.md — new ## Integrations section listing langchain-runcycles as the right fit for langchain.agents.create_agent workflows. The existing examples/langchain_integration.py row is reframed as the callback-handler path for non-agent runnables (bare ChatOpenAI, chains, RAG); LangChain 1.x agent middleware requires create_agent, so the two patterns serve different surfaces and both remain supported.
• examples/langchain_integration.py — file-level docstring updated to point agent-using readers at langchain-runcycles. Code unchanged.
• AUDIT.md — new entry "LangChain Agent Middleware Integration Pointer (added 2026-05-10)" documenting the docs-only nature of this change and the rationale for splitting agent middleware into a sibling package per LangChain's publishing guidance.
• .github/ISSUE_TEMPLATE/bug_report.yml, .github/ISSUE_TEMPLATE/feature_request.yml — add langchain-runcycles to the affected-repository dropdowns so users can file issues against the new package.

Why a sibling package

LangChain 1.x introduced an AgentMiddleware API with wrap_tool_call, before_model, and wrap_model_call hooks. The new package wraps that API on top of this SDK's existing decide / create_reservation / commit_reservation / release_reservation surface — no new SDK methods needed. Splitting into a sibling repo follows:

• LangChain's published guidance: "New integrations should be published as standalone PyPI packages."
• The langchain-<service> naming convention used by langchain-anthropic, langchain-openai, langchain-mongodb.

The new repo mirrors this one's CI / release / linter / coverage configuration so both repos remain operationally consistent.

Test plan

• README renders correctly on a local Markdown preview (Integrations section sits between Examples and Development; callback row reframed; new package linked)
• examples/langchain_integration.py still parses and runs (no code changes)
• Issue-template dropdowns updated in both bug_report.yml and feature_request.yml
• AUDIT.md entry follows the same template as prior entries (bold heading, file list, version, prose, protocol-conformance line at the end)