fix(address): normalize SECTORUL + etaj in addr_match, fix Oracle duplicate addrs

- _addr_match / addrMatch: add SECTORUL\s*\d* branch to strip sector
  number; add (?:\b|(?=\d)) to catch glued keywords (sc1, ap94);
  include etaj field in rStreet concat
- database.py: replace duplicate addr_match impl with import from sync_service
- import_service.py: short-circuit billing addr Oracle call when
  billing == shipping (avoids duplicate address creation)
- PL/SQL: normalize MUNICIPIUL BUCURESTI → BUCURESTI SECTORUL X before
  TIER 1; resolve id_localitate before search; TIER 1 now matches on
  id_loc instead of text locality
- Add scripts/cleanup_duplicate_addresses.sql for manual prod cleanup
- Add 5 new tests: sectorul, keyword+digit gluing, etaj, short-circuit

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

api/app/database.py

@@ -1,12 +1,8 @@
 import oracledb
 import aiosqlite
 import sqlite3
-import json
-import re
-import unicodedata
 import logging
 import os
-from pathlib import Path
 from .config import settings
 
 logger = logging.getLogger(__name__)
@@ -375,33 +371,7 @@ def init_sqlite():
 
 def _backfill_address_mismatch(conn):
     """Recompute address_mismatch from stored address JSON for all orders."""
-    _ADDR_WORDS = re.compile(
-        r'\b(STR|STRADA|NR|NUMAR|NUMARUL|BL|BLOC|SC|SCARA|AP|APART|APARTAMENT|'
-        r'ET|ETAJ|COM|COMUNA|SAT|MUN|MUNICIPIUL|JUD|JUDETUL|CARTIER|PARTER|SECTOR|ORAS)\b'
-    )
-
-    def norm(s):
-        s = unicodedata.normalize('NFD', s or '')
-        s = re.sub(r'[\u0300-\u036f]', '', s).upper()
-        s = _ADDR_WORDS.sub('', s)
-        return re.sub(r'[^A-Z0-9]', '', s)
-
-    def addr_match(gomag_json, roa_json):
-        if not gomag_json or not roa_json:
-            return True
-        try:
-            g = json.loads(gomag_json) if isinstance(gomag_json, str) else gomag_json
-            r = json.loads(roa_json) if isinstance(roa_json, str) else roa_json
-        except (json.JSONDecodeError, TypeError):
-            return True
-        g_street = norm(g.get('address') or g.get('strada') or '')
-        r_street = norm((r.get('strada') or '') + (r.get('numar') or ''))
-        g_city = norm(g.get('city') or g.get('localitate') or '')
-        r_city = norm(r.get('localitate') or '')
-        g_region = norm(g.get('region') or g.get('judet') or '')
-        r_region = norm(r.get('judet') or '')
-        return g_street == r_street and g_city == r_city and g_region == r_region
-
+    from .services.sync_service import _addr_match
     try:
         rows = conn.execute("""
             SELECT order_number, adresa_livrare_gomag, adresa_livrare_roa,
@@ -411,8 +381,8 @@ def _backfill_address_mismatch(conn):
         """).fetchall()
         updated = 0
         for r in rows:
-            livr_ok = addr_match(r[1], r[2])
-            fact_ok = addr_match(r[3], r[4])
+            livr_ok = _addr_match(r[1], r[2])
+            fact_ok = _addr_match(r[3], r[4])
             new_val = 1 if (not livr_ok or not fact_ok) else 0
             conn.execute(
                 "UPDATE orders SET address_mismatch = ? WHERE order_number = ?",