chore(rebar): Pier_T_rebar 작업 폴더 초기 구성 및 외부 의존 파일 통합

- Code\rebar_pipeline, review_pipeline, pier-program을 정션에서 실제 폴더로 전환
- Code\rhino_mcp 및 CLAUDE_rhino배경.md 추가, .mcp.json은 PYTHONPATH로 새 위치 사용
- pier-program 내부 .git 제거(이력은 backup\pier-program_20261028.bundle에 보관)
- worktree 가이드 반영(CLAUDE.local.md, HANDOFF.md)
- D:\Rhino-MCP 경로 참조를 새 위치로 치환

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

RAG/design-model-box-sample_2/tools/compare_saved_model_compatibility.py

@@ -0,0 +1,72 @@
+"""Compare baseline/staging classification against unchanged saved models.
+
+Read-only: no AI calls, job writes, geometry generation or approval changes.
+Run each checkout in a fresh interpreter to prevent mixed module imports.
+"""
+import argparse
+import copy
+import hashlib
+import json
+from pathlib import Path
+import subprocess
+import sys
+
+
+def digest(value):
+    return hashlib.sha256(json.dumps(value, sort_keys=True, separators=(',', ':')).encode()).hexdigest()
+
+
+def worker(root, job):
+    sys.path.insert(0, str(root / 'src/backend'))
+    from fresh_rebar_8097 import _compact_nonwing_rebar_taxonomy
+    raw = job.read_bytes()
+    plan = json.loads(raw)['result']
+    before = copy.deepcopy(plan)
+    groups = plan.get('rebar_groups') or []
+    _compact_nonwing_rebar_taxonomy(plan, groups)
+    fields = ('id', 'host_component_ids', 'instance_paths_m', 'diameter_mm')
+    physical = lambda rows: [{k: g.get(k) for k in fields} for g in rows]
+    assert physical(groups) == physical(before.get('rebar_groups') or []), 'rebar geometry/hosts changed'
+    assert plan.get('concrete_components') == before.get('concrete_components'), 'concrete changed'
+    assert job.read_bytes() == raw, 'saved job changed'
+    return dict(pass_=True, concrete=len(plan.get('concrete_components') or []),
+                groups=len(groups), paths=sum(len(g.get('instance_paths_m') or []) for g in groups),
+                physical_hash=digest(physical(groups)), concrete_hash=digest(plan.get('concrete_components')),
+                saved_job_sha256=hashlib.sha256(raw).hexdigest())
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('--baseline', type=Path)
+    parser.add_argument('--staging', type=Path)
+    parser.add_argument('--manifest', type=Path)
+    parser.add_argument('--jobs', type=Path)
+    parser.add_argument('--output', type=Path)
+    parser.add_argument('--worker-root', type=Path)
+    parser.add_argument('--worker-job', type=Path)
+    args = parser.parse_args()
+    if args.worker_root:
+        print(json.dumps(worker(args.worker_root, args.worker_job))); return
+    if args.output.exists():
+        parser.error('refusing to overwrite a comparison report')
+    rows = []
+    for item in json.loads(args.manifest.read_text())['models']:
+        row = dict(item)
+        for label, root in [('baseline', args.baseline), ('staging', args.staging)]:
+            result = subprocess.run([sys.executable, __file__, '--worker-root', str(root),
+                                     '--worker-job', str(args.jobs / (item['job_id'] + '.json'))],
+                                    capture_output=True, text=True)
+            row[label] = json.loads(result.stdout) if result.returncode == 0 else {
+                'pass_': False, 'error': result.stderr[-3000:]}
+        row['compatible'] = row['baseline'].get('pass_') and row['staging'] == row['baseline']
+        rows.append(row)
+        print(json.dumps({'name': item['name'], 'compatible': row['compatible'],
+                          'baseline': row['baseline'], 'staging': row['staging']}, ensure_ascii=False), flush=True)
+    with args.output.open('x') as stream:
+        json.dump({'scope': 'saved-model classification compatibility only; not regeneration or new engineering approval',
+                   'all_compatible': all(r['compatible'] for r in rows), 'results': rows}, stream, ensure_ascii=False, indent=2)
+    if not all(r['compatible'] for r in rows):
+        sys.exit(1)
+
+
+if __name__ == '__main__': main()