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/__init__.py

@@ -20,9 +20,11 @@ CONFIG = {}
 
 from .read import preprocess, extract_data_division, extract_procedure_division
 from .read import resolve_copybooks, parse_data_division, parse_file_section, scan_open_statements, parse_file_control
-from .core import build_branch_tree, classify_field_roles, _init_child_names
+from .core import classify_field_roles, _init_child_names
+from .pipeline_bridge import build_branch_tree_fallback
 from .cond import parse_single_condition, is_field, collect_leaves
-from .design import enum_paths, generate_records, _filter_stop
+from .design_mcdc import enum_paths, _filter_stop
+from .design import generate_records
 from .output import output_json, output_input_files
 from .coverage import run_coverage, generate_coverage_index, check_coverage
 from japanese_data import generate_fullwidth_text, generate_halfwidth_katakana, generate_wareki_date
@@ -249,7 +251,7 @@ def main():
         assignments = {}
 
         if proc_div:
-            branch_tree, assignments = build_branch_tree(proc_div, fields_dict)
+            branch_tree, assignments = build_branch_tree_fallback(proc_div, fields_dict)
 
             roles = classify_field_roles(branch_tree, assignments, fields_dict,
                                          source=preprocessed, proc_text=proc_div)
@@ -367,7 +369,7 @@ def extract_structure(cobol_source: str) -> dict:
     branch_tree = None
     assignments = {}
     if proc_div:
-        branch_tree, assignments = build_branch_tree(proc_div, fields_dict)
+        branch_tree, assignments = build_branch_tree_fallback(proc_div, fields_dict)
 
     file_sec = parse_file_section(preprocessed)
     open_dir = scan_open_statements(proc_div) if proc_div else {}
@@ -687,13 +689,38 @@ def generate_data(cobol_source: str, structure: dict = None) -> list[dict]:
 
     fields_dict = expand_occurs(fields_dict)
     proc_div = extract_procedure_division(preprocessed)
-    _, assignments = build_branch_tree(proc_div, fields_dict)
+    _, assignments = build_branch_tree_fallback(proc_div, fields_dict)
 
     file_sec = parse_file_section(preprocessed)
 
     branch_paths = enum_paths(branch_tree, fields_dict)
     branch_paths = [(_filter_stop(c), a) for c, a in branch_paths]
 
+    # Filter: remove constraints whose field doesn't exist in fields_dict.
+    # Resolve OF-qualified names and subscripts for matching.
+    _fdict_names = {f['name'] for f in fields_dict}
+    def _resolve_field(fn: str) -> str:
+        ufn = fn.upper()
+        if ' OF ' in ufn:
+            fn = fn.split(' OF ')[0].strip()
+        m = re.match(r'^(\w[\w-]*)\s*\(', fn)
+        if m and m.group(1) in _fdict_names:
+            return m.group(1)
+        return fn
+    filtered_paths = []
+    for cons_list, asgn in branch_paths:
+        clean = []
+        for c in cons_list:
+            if len(c) >= 4:
+                fn = _resolve_field(str(c[0]))
+                if fn in _fdict_names:
+                    c = list(c); c[0] = fn
+                    clean.append(tuple(c))
+            else:
+                clean.append(c)
+        filtered_paths.append((clean, asgn))
+    branch_paths = filtered_paths
+
     records, kept_paths = generate_records(branch_paths, fields_dict, assignments, file_sec=file_sec)
 
     # Cross-file KEY alignment for matching programs