Add invariant tests: coords, dedup, name fields, zh/ru quality

- test_all_coords_present: no null lat/lon
- test_no_duplicate_metro_rows: unique (lat, lon, name_en) — one row per metro
- test_all_name_fields_present: name_en, name_ru, name_zh all non-null
- test_name_zh_contains_cjk: name_zh must have CJK characters
- test_name_ru_has_cyrillic: name_ru must have Cyrillic
- test_no_typographic_quotes_in_ru: no «» in name_ru
- test_no_airport_words_in_zh: no 機場/空港 in name_zh
- test_balanced_parentheses: matched parens in all name fields

13/14 tests pass; test_lang_matches_region was already failing (pre-existing
ID/PH lang='zh' vs SINOPHONE_REGIONS definition gap).

test/test_invariants.py

@@ -9,9 +9,19 @@ Verifies:
 - name field matches lang selection (name_en/ru/zh)
 - lang matches region-based language rule
 - coordinates are in valid range
+- all name fields (name_en, name_ru, name_zh) are present
+- name_zh contains CJK characters (not Latin-only)
+- name_ru contains Cyrillic
+- name_ru has no typographic quotes «»
+- no airport words in name_zh
+- coordinates present and in valid range
+- no duplicate (lat, lon, name_en) groups
+- balanced parentheses in all name fields
 """
 
+import re
 import pytest
+from collections import defaultdict
 from datasets import load_dataset
 
 SLAVIC_REGIONS = {
@@ -94,3 +104,81 @@ def test_coordinates_valid(dataset):
            (e.get('lon') is not None and (e['lon'] < -180 or e['lon'] > 180))
     ]
     assert invalid == [], f"Invalid coordinates: {invalid[:5]}"
+
+
+def test_all_coords_present(dataset):
+    """Every row must have both lat and lon."""
+    missing = [
+        (e['code'], e.get('name_en'))
+        for e in dataset
+        if e.get('lat') is None or e.get('lon') is None
+    ]
+    assert missing == [], f"Null coordinates: {missing}"
+
+
+def test_no_duplicate_metro_rows(dataset):
+    """No two rows may share the same (lat, lon, name_en) — one row per metro."""
+    groups = defaultdict(list)
+    for e in dataset:
+        lat, lon, en = e.get('lat'), e.get('lon'), e.get('name_en') or ''
+        if lat is not None and lon is not None and en:
+            groups[(lat, lon, en)].append(e['code'])
+    dups = {k: v for k, v in groups.items() if len(v) > 1}
+    assert dups == {}, f"Duplicate (lat, lon, name_en) groups: {dict(list(dups.items())[:5])}"
+
+
+def test_all_name_fields_present(dataset):
+    """name_en, name_ru, and name_zh must all be non-null."""
+    for field in ('name_en', 'name_ru', 'name_zh'):
+        missing = [(e['code'], field) for e in dataset if not e.get(field)]
+        assert missing == [], f"Missing {field}: {missing[:10]}"
+
+
+def _has_cjk(s):
+    return any('一' <= c <= '鿿' or '㐀' <= c <= '䶿' for c in (s or ''))
+
+
+def _has_cyrillic(s):
+    return any('Ѐ' <= c <= 'ӿ' for c in (s or ''))
+
+
+def test_name_zh_contains_cjk(dataset):
+    """name_zh must contain CJK characters (not a Latin-only placeholder)."""
+    bad = [(e['code'], e['name_zh']) for e in dataset
+           if e.get('name_zh') and not _has_cjk(e['name_zh'])]
+    assert bad == [], f"name_zh with no CJK: {bad[:10]}"
+
+
+def test_name_ru_has_cyrillic(dataset):
+    """name_ru must contain Cyrillic characters."""
+    bad = [(e['code'], e['name_ru']) for e in dataset
+           if e.get('name_ru') and not _has_cyrillic(e['name_ru'])]
+    assert bad == [], f"name_ru without Cyrillic: {bad[:10]}"
+
+
+def test_no_typographic_quotes_in_ru(dataset):
+    """name_ru must not contain typographic quotes «»."""
+    bad = [(e['code'], e['name_ru']) for e in dataset
+           if e.get('name_ru') and re.search(r'[«»]', e['name_ru'])]
+    assert bad == [], f"name_ru with «» quotes: {bad[:10]}"
+
+
+_AIRPORT_ZH = re.compile(r'機場|机場|机场|空港')
+
+
+def test_no_airport_words_in_zh(dataset):
+    """name_zh must not contain airport-article words like 機場."""
+    bad = [(e['code'], e['name_zh']) for e in dataset
+           if e.get('name_zh') and _AIRPORT_ZH.search(e['name_zh'])]
+    assert bad == [], f"name_zh with airport word: {bad[:10]}"
+
+
+def test_balanced_parentheses(dataset):
+    """All name fields must have balanced ASCII and fullwidth parentheses."""
+    bad = []
+    for e in dataset:
+        for field in ('name', 'name_en', 'name_ru', 'name_zh'):
+            v = e.get(field) or ''
+            if v.count('(') != v.count(')') or v.count('（') != v.count('）'):
+                bad.append((e['code'], field, v))
+    assert bad == [], f"Unmatched parentheses: {bad[:5]}"