Restructure: descriptive tier and experiment names, paper/manuscript

- paper/pnas -> paper/manuscript (venue-neutral)
- configs/layer1 -> configs/inheritance, src/knowledge -> src/inheritance
  (imported as `inheritance`), make layer1 -> make inheritance; layer2 alias dropped
- inheritance and trained-network bundles named after the manuscript figure
  they feed (fig2_grounding_sweep, figS3_rebaselining, ...), or descriptively
  where they feed none; configs keep their `experiment:` value so parquet
  hashes are unchanged, only output.dir moves
- figure scripts, SI figure sources, notebooks, REPRODUCING.md, README and the
  SI Methods/tables updated; make clean no longer deletes tracked manifests;
  reproduce.sh hashes the s{seed}/ layouts too

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Y64o8FKP7rCuXzC48pxpMm

tests/test_neural_validation.py

@@ -4,7 +4,7 @@ With ``model.kind == "histogram"`` the neural generational step (train-on-parent
 + grounding) is *exactly* neutral Wright-Fisher drift with immigration. This module asserts
 that the neural runner reproduces the two closed forms Layer 1 is validated against
 (blueprint 2.4-1 neutral heterozygosity decay, 2.4-3 exact mutation-drift equilibrium) and
-that its mean H-trajectory tracks ``knowledge.lineage.run_lineage`` directly. If any of
+that its mean H-trajectory tracks ``inheritance.lineage.run_lineage`` directly. If any of
 these fail the neural plumbing is wrong — no real network should be trained until they pass.
 """
 
@@ -13,8 +13,8 @@ from __future__ import annotations
 import numpy as np
 import pytest
 
-from knowledge.lineage import run_lineage
-from knowledge.seeding import spawn_seeds
+from inheritance.lineage import run_lineage
+from inheritance.seeding import spawn_seeds
 from neural.generation_loop import run_generative_lineage
 from neural.synthetic import make_mode_truth
 from neural.config import SyntheticCfg