Add repair RL autogen pipeline

[Peter7896]

Summary

• add a dependency-free RL autogen CLI that converts SWE-style issues plus Agentless patch/test outputs into rollout and reward manifest JSONL
• add fixture-driven unittest coverage for deduplication, score filtering, and summary output
• document how this complements the existing evaluation pipeline and issue please release swe-enviroment #1 online RL discussion

Scope

This targets the SWE-Swiss bounty's full auto-generation path by adding a tested conversion layer between public repair/evaluation artifacts and downstream RL training jobs. It does not claim to run large-scale training or Docker reward execution in this PR.

Validation

• python -m py_compile RL\autogen_pipeline.py RL\tests\test_autogen_pipeline.py
• python -m unittest discover RL\tests
• git diff --cached --check

Refs #1

/claim https://algora.io/PrimeIntellect-ai/bounties/FHr4fCjVPjFDmFED