fix: HINA 全类型缺陷修复 — SORT/CSV/ALT 3 个真实缺陷

对抗性全类型测试发现的缺陷和修复:

缺陷1: SORT/MERGE L1 关键词太严格（漏检）
  - 旧: 'SORT ON KEY' / 'MERGE ON KEY'（精确字符串）
  - COBOL 中的真实写法: SORT WORK-FILE ON ASCENDING KEY ...
  - 新: 正则 SORT(?:\s+\S+)?\s+ON\s+(?:ASCENDING|DESCENDING)?KEY

缺陷2: CSV 假阳性（STRING/INSPECT 非CSV也触发）
  - 旧: has_string=True -> CSV合并
  - 新: 要求 has_csv_merge（STRING+逗号分隔）
  - 单纯字符串拼接不再触发 CSV 分类

缺陷3: ALTERNATE RECORD KEY 被 ORGANIZATION IS 覆盖
  - 旧: 文件编成先于替代索引（同确信度先者胜）
  - 新: 替代索引放前面（更具体的分类优先）

回归: 767 passed（0 new failures）

tests/hina/test_classifier_deep.py

@@ -20,7 +20,7 @@ def test_detect_keyword_multiple_matches():
            EXEC SQL
                SELECT * FROM TABLE
            END-EXEC.
-           SORT ON KEY WS-KEY.
+           SORT SORT-FILE ON KEY WS-KEY.
            CALL 'SUBPGM'.
            STOP RUN.
     """
@@ -36,7 +36,7 @@ def test_detect_keyword_multiple_matches():
     assert cat_map["DB操作"][0] == 0.95
     assert cat_map["DB操作"][1] == "EXEC SQL"
     assert cat_map["SORT"][0] == 0.95
-    assert cat_map["SORT"][1] == "SORT ON KEY"
+    assert cat_map["SORT"][1].startswith("re:SORT")  # regex pattern
     assert cat_map["子程序调用"][0] == 0.90
     assert cat_map["子程序调用"][1] == "CALL"
 
@@ -154,7 +154,7 @@ def test_detect_keyword_mixed_case_whitespace_comments():
     matched_keywords = {r[2] for r in results}
     assert "EXEC SQL" in matched_keywords
     assert "CALL" in matched_keywords
-    assert "SORT ON KEY" in matched_keywords
+    assert any(r[0] == "SORT" for r in results)  # SORT detected via regex
 
 
 # ── 5. No keyword match and no LLM result → unknown ──
@@ -190,8 +190,8 @@ def test_detect_keyword_all_rules():
         ("ALPHABETIC", "编码转换"),
         ("DFHCOMMAREA", "online"),
         ("MAP", "online"),
-        ("SORT ON KEY", "SORT"),
-        ("MERGE ON KEY", "MERGE"),
+        ("SORT SORT-FILE ON KEY", "SORT"),
+        ("MERGE MERGE-FILE ON KEY", "MERGE"),
         ("WRITE AFTER", "编辑输出"),
         ("WRITE BEFORE", "编辑输出"),
         ("ORGANIZATION IS", "文件编成"),