refactor(pdf_extractor): 移除 toc_detector 导入保护，强制依赖生效

- 去掉 try-except 包裹的延迟导入
- 导入失败时直接抛异常，确保依赖问题及时暴露

core/construction_review/component/minimal_pipeline/pdf_extractor.py

@@ -148,12 +148,8 @@ class PdfStructureExtractor:
         Returns:
             {"chapters": [...], "total_chapters": N} 或 None
         """
-        # 延迟导入避免循环依赖
-        try:
-            from .toc_detector import TOCCatalogExtractor
-        except ImportError:
-            logger.warning("[PDF提取] toc_detector 模块未找到，跳过目录检测")
-            return None
+        # 延迟导入避免循环依赖（YOLO依赖必须存在，否则报错）
+        from .toc_detector import TOCCatalogExtractor
 
         if self._toc_extractor is None:
             self._toc_extractor = TOCCatalogExtractor(