F1=62.7%: Full iter 8 results, random eval, WarpGrep description cleanup

- Add --random flag for 13-PR eval (2 per source_repo, seed=42)
- Simplify WarpGrep tool description (concise, actionable)
- Update plan.md with full score history through iter 8
- Latest eval: P=87.5%, R=48.8%, F1=62.7% (21 TP, 6 FP, 22 FN)