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/flatfile.py

@@ -94,10 +94,16 @@ def _format_value(value: Any, field: dict) -> bytes:
 
 
 def write_flat_file(records: list[dict], layout: dict, outpath: Path, field_filter: set = None):
-    """Write records as a COBOL-compatible fixed-length flat file."""
+    """Write records as a COBOL-compatible fixed-length flat file.
+
+    Supports multi-record FDs: uses the longest record layout (most fields)
+    to maximize compatible field coverage.
+    """
+    outpath = Path(outpath)
     if not layout or not layout.get("records"):
         return
-    rec = layout["records"][0]
+    # Pick the record with the most fields (best coverage for multi-record FDs)
+    rec = max(layout["records"], key=lambda r: (len(r["fields"]), r["record_length"]))
     rec_len = rec["record_length"]
     if rec_len == 0:
         return
@@ -119,6 +125,7 @@ def write_flat_file(records: list[dict], layout: dict, outpath: Path, field_filt
 
 def write_all_files(records: list[dict], source_text: str, outdir: Path, prefix: str = ""):
     """Analyze source, write flat files for all INPUT FDs."""
+    outdir = Path(outdir)
     layouts = analyze_fd_layout(source_text)
     written = []
     for filename, layout in layouts.items():
@@ -128,9 +135,20 @@ def write_all_files(records: list[dict], source_text: str, outdir: Path, prefix:
         for rec in layout["records"]:
             for f in rec["fields"]:
                 fnames.add(f["name"])
+        if not fnames:
+            continue
+        # Filter generated records to only include fields from this FD
         filtered = [{k: v for k, v in r.items() if k in fnames} for r in records]
-        if filtered and any(v for row in filtered for v in row.values()):
+        has_data = any(v for row in filtered for v in row.values())
+        if not has_data:
+            # Fallback: one zero-filled record from FD layout
+            fallback = {}
+            for rec in layout["records"]:
+                for f in rec["fields"]:
+                    fallback[f["name"]] = 0 if f["type"] == "numeric" else " "
+            filtered = [fallback] if fallback else []
+        if filtered:
             outpath = outdir / (prefix + filename)
-            write_flat_file(records, layout, outpath)
-            written.append((filename, outpath, len([r for r in filtered if any(v for v in r.values())])))
+            write_flat_file(filtered, layout, outpath)
+            written.append((filename, outpath, len(filtered)))
     return written