refactor: split line_items_extractor into smaller modules with comprehensive tests

- Extract models.py (LineItem, LineItemsResult dataclasses)
- Extract html_table_parser.py (ColumnMapper, HtmlTableParser)
- Extract merged_cell_handler.py (MergedCellHandler for PP-StructureV3 merged cells)
- Reduce line_items_extractor.py from 971 to 396 lines
- Add constants for magic numbers (MIN_AMOUNT_THRESHOLD, ROW_GROUPING_THRESHOLD, etc.)
- Fix row grouping algorithm in text_line_items_extractor.py
- Demote INFO logs to DEBUG level in structure_detector.py
- Add 209 tests achieving 85%+ coverage on main modules

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

tests/table/test_text_line_items_extractor.py

@@ -142,6 +142,33 @@ class TestTextLineItemsExtractor:
         rows = extractor._group_by_row(elements)
         assert len(rows) == 2
 
+    def test_group_by_row_varying_heights_uses_average(self, extractor):
+        """Test grouping handles varying element heights using dynamic average.
+
+        When elements have varying heights, the row center should be recalculated
+        as new elements are added, preventing tall elements from being incorrectly
+        grouped with the next row.
+        """
+        # First element: small height, center_y = 105
+        # Second element: tall, center_y = 115 (but should still be same row)
+        # Third element: next row, center_y = 160
+        elements = [
+            TextElement(text="Short", bbox=(0, 100, 100, 110)),      # center_y = 105
+            TextElement(text="Tall item", bbox=(150, 100, 250, 130)),  # center_y = 115
+            TextElement(text="Next row", bbox=(0, 150, 100, 170)),   # center_y = 160
+        ]
+        rows = extractor._group_by_row(elements)
+
+        # With dynamic average, both first and second element should be same row
+        assert len(rows) == 2
+        assert len(rows[0]) == 2  # Short and Tall item
+        assert len(rows[1]) == 1  # Next row
+
+    def test_group_by_row_empty_input(self, extractor):
+        """Test grouping with empty input returns empty list."""
+        rows = extractor._group_by_row([])
+        assert rows == []
+
     def test_looks_like_line_item_with_amount(self, extractor):
         """Test line item detection with amount."""
         row = [
@@ -253,6 +280,67 @@ class TestTextLineItemsExtractor:
         assert len(elements) == 4
 
 
+class TestExceptionHandling:
+    """Tests for exception handling in text element extraction."""
+
+    def test_extract_text_elements_handles_missing_bbox(self):
+        """Test that missing bbox is handled gracefully."""
+        extractor = TextLineItemsExtractor()
+        parsing_res = [
+            {"label": "text", "text": "No bbox"},  # Missing bbox
+            {"label": "text", "bbox": [0, 100, 200, 120], "text": "Valid"},
+        ]
+        elements = extractor._extract_text_elements(parsing_res)
+        # Should only have 1 valid element
+        assert len(elements) == 1
+        assert elements[0].text == "Valid"
+
+    def test_extract_text_elements_handles_invalid_bbox(self):
+        """Test that invalid bbox (less than 4 values) is handled."""
+        extractor = TextLineItemsExtractor()
+        parsing_res = [
+            {"label": "text", "bbox": [0, 100], "text": "Invalid bbox"},  # Only 2 values
+            {"label": "text", "bbox": [0, 100, 200, 120], "text": "Valid"},
+        ]
+        elements = extractor._extract_text_elements(parsing_res)
+        assert len(elements) == 1
+        assert elements[0].text == "Valid"
+
+    def test_extract_text_elements_handles_none_text(self):
+        """Test that None text is handled."""
+        extractor = TextLineItemsExtractor()
+        parsing_res = [
+            {"label": "text", "bbox": [0, 100, 200, 120], "text": None},
+            {"label": "text", "bbox": [0, 150, 200, 170], "text": "Valid"},
+        ]
+        elements = extractor._extract_text_elements(parsing_res)
+        assert len(elements) == 1
+        assert elements[0].text == "Valid"
+
+    def test_extract_text_elements_handles_empty_string(self):
+        """Test that empty string text is skipped."""
+        extractor = TextLineItemsExtractor()
+        parsing_res = [
+            {"label": "text", "bbox": [0, 100, 200, 120], "text": ""},
+            {"label": "text", "bbox": [0, 150, 200, 170], "text": "Valid"},
+        ]
+        elements = extractor._extract_text_elements(parsing_res)
+        assert len(elements) == 1
+        assert elements[0].text == "Valid"
+
+    def test_extract_text_elements_handles_malformed_element(self):
+        """Test that completely malformed elements are handled."""
+        extractor = TextLineItemsExtractor()
+        parsing_res = [
+            "not a dict",  # String instead of dict
+            123,  # Number instead of dict
+            {"label": "text", "bbox": [0, 100, 200, 120], "text": "Valid"},
+        ]
+        elements = extractor._extract_text_elements(parsing_res)
+        assert len(elements) == 1
+        assert elements[0].text == "Valid"
+
+
 class TestConvertTextLineItem:
     """Tests for convert_text_line_item function."""