results: rmse outlier-KL barrier holds coherence over the loop; README + log-incoherence plot

Headline (gemma-3-4b-it s42, care-over-authority): aggregating the kl_rev
barrier by rmse over token positions (not the mean) holds coherence flat at
0.997 across all 8 rounds, where the mean aggregate collapses to 0.62 by r7
(token loops). Mean dilutes the few incoherent positions under the tau gate;
rmse is outlier-sensitive and fires on them. Cost is depth (rmse run leashes
to base, trait stays shallow); matched control still running.

- plot.py: coherence panel -> log-incoherence (1-coh, log axis, down=coherent);
  map coherence axis matches; red steer kept on the over-pipeline panels only.
- heal.py: fix kl_agg=p95 crash (torch.quantile rejects bf16 -> .float()).
- run.py: persist per-round adapter gens (adapter_gen) for the outputs table.
- config.py: coh_floor early-stop knob.
- README: results table (mean vs rmse), trajectory figure, outputs-over-loop
  appendix (per-round completions as quotes); spec persona corrected to pos-neg.
- docs/reviews: kl_agg review, pool saturation test, care-lens plan.

Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>

scripts/plot_run.py

@@ -16,7 +16,7 @@ from pathlib import Path
 import srsly
 import tyro
 
-from steer_heal.plot import write_trajectory
+from steer_heal.plot import write_report, write_trajectory
 
 
 def main(run_dir: Path, base_auth: float | None = None,
@@ -49,6 +49,11 @@ def main(run_dir: Path, base_auth: float | None = None,
     png = write_trajectory(run_dir, stages)
     print(f"re-rendered {png} from {len(stages)} stages ({len(healed)} rounds)", file=sys.stderr)
 
+    gen_rounds = sorted(by_stage("adapter_gen"), key=lambda e: e["round"])
+    if gen_rounds:  # runs from before adapter_gen was persisted (e.g. #107/#108) have none
+        out = write_report(run_dir, gen_rounds)
+        print(f"re-rendered {out} from {len(gen_rounds)} rounds of gens", file=sys.stderr)
+
 
 if __name__ == "__main__":
     tyro.cli(main)