OPCIÓN C + A1: Consolidación técnica + orjson

C1: Materialized view part_vehicle_preview (creación en progreso)
- Migración v3.3_materialized_view.sql
- catalog_service.py y dashboard/server.py ahora usan la MV
- Script refresh_part_vehicle_preview.py + warm_vehicle_cache.py actualizado

C2: Fix cache warming script (autónomo)
- Auto-re-ejecuta con sudo -u postgres si peer auth falla
- Args CLI: --dsn, --batch-size, --ttl, --dry-run

C3: CSS dinámico residual extraído
- sidebar.js → sidebar.css (nuevo)
- pos-utils.js → common.css (nuevo)
- Links agregados a 14 templates POS

C4: Script de load testing básico
- scripts/load_test.py: métricas p50/p95/p99, throughput, errores

C5: Documentación actualizada
- FASES_IMPLEMENTADAS.md: test count real, FASE 7 completa
- performance_audit_2026.md: anexo post-FASE 7, métricas actualizadas

A1: Serialización orjson
- pos/json_provider.py: DefaultJSONProvider con orjson.dumps/loads
- Aplicado a POS app y Dashboard server
- Fix indentation error en pos_bp.py

Tests: 73/73 pasando

pos/services/catalog_service.py

@@ -1372,15 +1372,9 @@ def smart_search(master_conn, q, tenant_conn, branch_id, limit=50):
 
     if missing_ids:
         cur.execute("""
-            SELECT DISTINCT ON (vp.part_id)
-                   vp.part_id, b.name_brand, m.name_model, y.year_car
-            FROM vehicle_parts vp
-            JOIN model_year_engine mye ON mye.id_mye = vp.model_year_engine_id
-            JOIN models m ON m.id_model = mye.model_id
-            JOIN brands b ON b.id_brand = m.brand_id
-            JOIN years y ON y.id_year = mye.year_id
-            WHERE vp.part_id = ANY(%s)
-            ORDER BY vp.part_id, y.year_car DESC
+            SELECT part_id, name_brand, name_model, year_car
+            FROM part_vehicle_preview
+            WHERE part_id = ANY(%s)
         """, (missing_ids,))
         for row in cur.fetchall():
             info = f"{row[1]} {row[2]} {row[3]}"