Enable end-to-end LeanCI PR review pipeline

[CodewithJha]

Summary

• Wire the GitHub Action through ParitokGateway and the orchestrator so PR runs execute a real review instead of dry-run config printing.
• Add LLM client, agent loop, findings normalizer, pricing/receipt, GitHub publisher, metrics export, and dual-run controller (gated behind mode=dual_run).
• Document required secrets and the first end-to-end verification checklist in the README.

Test plan

• Confirm LeanCI workflow runs on this PR
• Confirm Paritok proxy starts successfully
• Confirm review executes and posts/updates a PR comment with findings + cost receipt
• Confirm leanci-metrics artifact uploads
• Confirm a follow-up push updates the same comment (no duplicates)
• Local: pytest (254 tests) green

[github-actions]

LeanCI could not complete

Error class: LLMError
What happened: LLM HTTP 400 from http://127.0.0.1:8080/v1/chat/completions: Bad Request — {"error":{"message":"Failed to call a function. Please adjust your prompt. See 'failed_generation' for more details.","type":"invalid_request_error","code":"tool_use_failed","failed_generation":"<function=expand_context[]{"shadow_id": "e07d01a14ceaf205"}</function>"}}
Next steps:

• Check OPENAI_API_KEY and provider rate limits
• Re-run after a short backoff
  Run: https://github.com/CodewithJha/leanci/actions/runs/30612579049