fix(F-09): address review-gate findings — resource leak + idempotency

- mathpix.py + figures.py: wrap fitz page-loop in try/finally so
  doc.close() is guaranteed even on exception (no file-handle leak)
- ingest.py: DELETE FROM formulas/figures WHERE paper_id before re-insert
  so --rich re-ingest is idempotent (BIGSERIAL has no natural UNIQUE key;
  ON CONFLICT DO NOTHING was a no-op)

Gate: 158 passed, ruff clean, mypy clean

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

codex/parsing/mathpix.py

@@ -112,54 +112,60 @@ def _extract_formulas_pix2tex(pdf_path: str) -> list[FormulaChunk]:
 
     scale = fitz.Matrix(3, 3)
 
-    for page_num in range(len(doc)):
-        page = doc[page_num]
-        blocks: list[Any] = page.get_text("blocks")
+    try:
+        for page_num in range(len(doc)):
+            page = doc[page_num]
+            blocks: list[Any] = page.get_text("blocks")
 
-        for block in blocks:
-            # block = (x0, y0, x1, y1, text, block_no, block_type)
-            if len(block) < 5:
-                continue
-            x0, y0, x1, y1 = block[0], block[1], block[2], block[3]
-            text: str = block[4]
+            for block in blocks:
+                # block = (x0, y0, x1, y1, text, block_no, block_type)
+                if len(block) < 5:
+                    continue
+                x0, y0, x1, y1 = block[0], block[1], block[2], block[3]
+                text: str = block[4]
 
-            # Size filter
-            width_pt = x1 - x0
-            height_pt = y1 - y0
-            if width_pt < _MIN_WIDTH_PT or height_pt < _MIN_HEIGHT_PT:
-                continue
+                # Size filter
+                width_pt = x1 - x0
+                height_pt = y1 - y0
+                if width_pt < _MIN_WIDTH_PT or height_pt < _MIN_HEIGHT_PT:
+                    continue
 
-            # Math content filter
-            if _math_symbol_count(text) < _MIN_MATH_SYMBOLS or _math_ratio(text) < _MIN_MATH_RATIO:
-                continue
+                # Math content filter
+                if (
+                    _math_symbol_count(text) < _MIN_MATH_SYMBOLS
+                    or _math_ratio(text) < _MIN_MATH_RATIO
+                ):
+                    continue
 
-            # Crop at 3× for OCR quality
-            clip_rect = fitz.Rect(x0, y0, x1, y1)
-            crop_pix = page.get_pixmap(matrix=scale, clip=clip_rect)
-            img = Image.frombytes(
-                "RGB",
-                (crop_pix.width, crop_pix.height),
-                crop_pix.samples,
-            )
-
-            try:
-                latex: str = model(img)
-            except Exception:
-                logger.debug("pix2tex failed on page %d block, skipping", page_num + 1)
-                continue
-
-            if not latex or not latex.strip():
-                continue
-
-            context = text[:200].replace("\n", " ")
-            results.append(
-                FormulaChunk(
-                    paper_id=paper_id,
-                    page=page_num + 1,
-                    raw_latex=latex.strip(),
-                    context=context,
+                # Crop at 3× for OCR quality
+                clip_rect = fitz.Rect(x0, y0, x1, y1)
+                crop_pix = page.get_pixmap(matrix=scale, clip=clip_rect)
+                img = Image.frombytes(
+                    "RGB",
+                    (crop_pix.width, crop_pix.height),
+                    crop_pix.samples,
                 )
-            )
+
+                try:
+                    latex: str = model(img)
+                except Exception:
+                    logger.debug("pix2tex failed on page %d block, skipping", page_num + 1)
+                    continue
+
+                if not latex or not latex.strip():
+                    continue
+
+                context = text[:200].replace("\n", " ")
+                results.append(
+                    FormulaChunk(
+                        paper_id=paper_id,
+                        page=page_num + 1,
+                        raw_latex=latex.strip(),
+                        context=context,
+                    )
+                )
+    finally:
+        doc.close()
 
     return results