Feature/langchain e2e memory benchmark

[boldirev-as]

Summary

Type of change

• Bug fix
• New feature / detector / policy primitive
• Framework or backend adapter
• Docs / examples
• Refactor / internal

Testing

• pytest passes locally
• Added or updated tests for the change
• If docs changed, links and code samples render correctly

Checklist

• Follows the conventions in CONTRIBUTING.md
• No secrets, credentials, or PII committed
• No unverifiable claims added to README or docs

[vgudur-dev]

Thanks for the comprehensive benchmark work, @boldirev-as! This is a substantial contribution.

Before we can merge, a few items to address:

1. Merge conflict — please rebase onto main and resolve the README.md conflict.
2. 2. Documentation language — the project's primary language is English. Could you translate the Russian documentation (commit c86ca23, 04e8acc) to English? A Russian translation can remain alongside as a secondary file if you'd like.
3. 3. CI checks — I've approved the workflows; they're running now. Please ensure tests pass.
      Once conflicts are resolved and docs are in English, happy to merge. Great work on the multi-step poisoning scenarios!