提升：37/37基准程序全量解析+O(N)路径枚举+运行时gcov验证

## 核心变更

### 1. 新PROCEDURE DIVISION解析器（procedure_parser.py）
- 行级状态机替换旧的BrParser regex解析器
- 覆盖：IF/ELSE/END-IF（嵌套）、EVALUATE/WHEN/ALSO、
  PERFORM UNTIL/VARYING、READ/AT END/NOT AT END、
  SORT/MERGE、GO TO DEPENDING ON
- 之前：3/37程序有分支检测  →  现在：37/37全部有分支
- 速度：~20ms/程序，纯规则引擎

### 2. 桥接层（pipeline_bridge.py）
- 新解析器为主，旧解析器3秒超时兜底
- 自动选取分支数更多的结果

### 3. 线性路径枚举（design_mcdc.py）
- 替换旧的Cartesian积路径枚举（O(2^N)）为每决策点独立枚举（O(N)）
- 28-sysin: 162分支仅163条路径（之前需截断到60DP）
- 消除了500路径硬上限和60DP截断

### 4. 条件解析修复（cond.py）
- NOT运算符规范化：X NOT = 5 → X <> 5
- 88-level反向：NOT WS-EOF-Y → parent <> value
- 裸字段引用：NOT WS-EOF → WS-EOF <> 'Y'
- 验证：1182个IF条件中0个NOT污染

### 5. 约束字段过滤（__init__.py）
- OF限定词剥离：STD-KEY OF MASTER-REC → STD-KEY
- 下标字段解析：WS-ITEM(SUB) → WS-ITEM
- 跳过不在fields_dict中的字段（group item/伪影）

### 6. 预处理器增强（read.py）
- VALUE ALL剥离（VALUE ALL '*' → VALUE '*'）
- &续行合并（COBOL多行字符串拼接）
- PIC小数点点→V转换（Z(9)9.99. → Z(9)9V99.）
- 缺少点号补全

### 7. Grammar修复（grammar.lark）
- OCCURS 1 TIME支持（原只认TIMES）
- USAGE IS COMP支持（可选IS）
- $符号在PICTURE_STRING中
- 无NAME条款支持（clause+）

### 8. Flatfile写入（flatfile.py）
- 多记录FD支持（选字段最多的记录）
- Path类型强制转换
- 回退零值记录

### 9. Bug修复
- trace_to_root空列表保护（core.py）

### 10. 测试套件（S16-S21）
- S16: 全量基准程序端到端
- S17: gcov运行时对比
- S18/S19: 桥接器验证
- S20: DISPLAY插桩运行时验证+gcov分支覆盖率
- S21: 条件解析修复验证
- 全部17/17回归测试通过

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

cobol_testgen/read.py

@@ -43,6 +43,31 @@ def preprocess(source: str) -> str:
         return re.sub(r'\s*,\s*', ' ', m.group(0))
     source = re.sub(r'VALUE\s+[^.\n]+', _strip_value_commas, source, flags=re.IGNORECASE)
 
+    # Strip ALL from VALUE ALL (VALUE ALL '*.' → VALUE '*.')
+    source = re.sub(r'\bVALUE\s+ALL\b', 'VALUE', source, flags=re.IGNORECASE)
+
+    # Collapse &-concatenated VALUE continuation lines
+    # COBOL uses & to split long literals across lines:
+    #   "............................" &
+    #   "............................"
+    # Match: (quote/X'...') + " &" + newline + (quote/X'...')
+    source = re.sub(
+        r'([Xx]?["\'])\s*&\s*\n\s*([Xx]?["\'])',
+        lambda m: m.group(1) + m.group(2),
+        source
+    )
+
+    # Remove trailing & at end of lines (standalone continuation markers)
+    source = re.sub(r'&(?=[^"\']*$)', '', source, flags=re.MULTILINE)
+
+    # Convert PIC decimal dots to V (implied decimal) for Lark compatibility
+    # PIC Z(9)9.99. → PIC Z(9)9V99.  (only within PIC clause before DOT)
+    source = re.sub(
+        r'(PIC\s+)([A-Z0-9(),\-*/V\$]+)\.(\d+)',
+        r'\1\2V\3',
+        source, flags=re.IGNORECASE
+    )
+
     fixed = _is_fixed_format(source)
     lines = []
     for raw_line in source.splitlines():
@@ -67,9 +92,25 @@ def preprocess(source: str) -> str:
             line = line.strip()
             if not line:
                 continue
+            # Strip bare * comment lines in free format (after *> removal)
+            if line.startswith('*') and not line.startswith('*>'):
+                continue
             content = line
         lines.append(re.sub(r'\s+FALSE\s+[^\s.]+', '', content.upper()))
-    return '\n'.join(lines)
+
+    # Ensure DATA DIVISION lines with PIC/VALUE but no trailing DOT get one
+    # (handles COBOL programs where the period on a PIC clause is optional/omitted)
+    fixed_lines = []
+    for i, line in enumerate(lines):
+        stripped = line.strip()
+        if stripped and not stripped.endswith('.'):
+            # Lines inside DATA DIVISION that have PIC or VALUE but no DOT
+            if re.search(r'\b(PIC|VALUE|REDEFINES|OCCURS|USAGE)\b', stripped, re.IGNORECASE):
+                # Only fix if the NEXT line also looks like a data_item (level_num)
+                if i + 1 < len(lines) and re.match(r'^\s*(0[1-9]|[0-4][0-9]|49|66|77|88)\s', lines[i + 1]):
+                    line = line.rstrip() + ' .'
+        fixed_lines.append(line)
+    return '\n'.join(fixed_lines)
 
 
 def extract_data_division(source: str) -> str:
@@ -97,13 +138,18 @@ def extract_procedure_division(source: str) -> str:
 _COPYBOOK_EXTENSIONS = ['.cpy', '.cbl', '.cpb', '']
 
 
-def resolve_copybooks(source: str, source_dir: str, _recursion_depth: int = 0) -> str:
-    """Find COPY statements and replace with copybook content."""
+def resolve_copybooks(source: str, source_dir: str, _recursion_depth: int = 0,
+                       extra_search_paths: list[str] = None) -> str:
+    """Find COPY statements and replace with copybook content.
+
+    Searches from source_dir first, then extra_search_paths.
+    """
     _RE_COPY = re.compile(
         r"^\s*COPY\s+(\w[\w-]*|\"[^\"]*\"|\'[^\']*\')(?:\s+REPLACING\s+(.+?))?\s*\.?\s*$",
         re.IGNORECASE
     )
     _RE_PAIR = re.compile(r"==(.+?)==\s+BY\s+==(.+?)==", re.IGNORECASE)
+    search_dirs = [source_dir] + (extra_search_paths or [])
 
     lines = source.split('\n')
     result = []
@@ -113,10 +159,13 @@ def resolve_copybooks(source: str, source_dir: str, _recursion_depth: int = 0) -
             raw_name = m.group(1)
             name = raw_name.strip('"').strip("'").upper()
             found = None
-            for ext in _COPYBOOK_EXTENSIONS:
-                p = Path(source_dir, name + ext)
-                if p.exists():
-                    found = p
+            for sd in search_dirs:
+                for ext in _COPYBOOK_EXTENSIONS:
+                    p = Path(sd, name + ext)
+                    if p.exists():
+                        found = p
+                        break
+                if found:
                     break
             if found:
                 if _recursion_depth > 10: