Implement proposal #19: Score-and-Retry + JSON format validation for 知微 pipeline

- Create json_validator.py utility module:
  - validate_json_syntax(): JSON syntax scoring (0-100) with auto-fix
  - validate_with_schema(): Multi-layer validation (syntax + schema)
  - score_and_retry(): Generic retry wrapper with error feedback
  - safe_parse_structured(): Safe <structured_data> extraction
  - FormatErrorLibrary: Persistent error tracking + stats + auto recommendations
  - CLI entry point for standalone use

- Integrate into strategy_lifecycle.py:
  - safe_json_load() now uses json_validator for auto-fix + error logging
  - Falls back to legacy repair on ImportError

- Integrate into update_data.py:
  - parse_report() uses safe_parse_structured() instead of bare json.loads()
  - Errors logged to FormatErrorLibrary silently

- Convert strategy_lifecycle.py to Unix line endings

update_data.py

@@ -55,20 +55,32 @@ def parse_report(markdown_text, source_file=None):
         report["type"] = "盘前"
 
     # ★ 优先提取结构化JSON（如果知微输出了的话）
-    struct_match = re.search(r'<structured_data>\s*(\{.*?\})\s*</structured_data>', markdown_text, re.DOTALL)
+    struct_match = re.search(r'<structured_data>\s*(.*?)\s*</structured_data>', markdown_text, re.DOTALL)
     if struct_match:
         try:
-            parsed = json.loads(struct_match.group(1))
-            report["structured"] = parsed
-            # 从结构化数据中直接取stock codes
-            codes = set()
-            for h in parsed.get("holdings", []):
-                c = h.get("code", "")
-                if c:
-                    codes.add(c)
-            report["stocks_mentioned"] = sorted(codes)
-        except (json.JSONDecodeError, Exception) as e:
-            pass  # JSON解析失败→走NLP兜底
+            import sys as _sys
+            _sys.path.insert(0, "/home/hmo/MoFin")
+            from json_validator import safe_parse_structured
+
+            parsed_sd, sd_score, sd_errors = safe_parse_structured(
+                struct_match.group(0),
+                log_errors=True,
+            )
+            if parsed_sd:
+                report["structured"] = parsed_sd
+                report["_json_score"] = sd_score
+                # 从结构化数据中直接取stock codes
+                codes = set()
+                for h in parsed_sd.get("holdings", []):
+                    c = h.get("code", "")
+                    if c:
+                        codes.add(c)
+                report["stocks_mentioned"] = sorted(codes)
+            elif sd_errors:
+                # 记录错误但继续，不阻断流程
+                pass
+        except Exception:
+            pass  # 兜底，不阻断 pipeline
 
     # 摘要（前3非空行）
     body_lines = [l.strip() for l in lines if l.strip() and not l.strip().startswith("#") and not l.strip().startswith("##")]