Fix page_ref extraction: only extract cells matching page-ref pattern

Column_1 cells like "to" (infinitive markers) were incorrectly extracted
as page_refs. Now only cells matching p.70, ,.65, or bare digits are
treated as page references.

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

klausur-service/backend/grid_editor_api.py

@@ -1884,6 +1884,8 @@ async def _build_grid_core(session_id: str, session: dict) -> dict:
     # = page number at bottom) are standalone rows that should be
     # removed from the table entirely.
     _REAL_IPA_CHARS_SET = set("ˈˌəɪɛɒʊʌæɑɔʃʒθðŋ")
+    # Page-ref pattern: "p.70", "P.70", ",.65" (garbled "p"), or bare "70"
+    _PAGE_REF_RE = re.compile(r'^[pP,]?\s*\.?\s*\d+$')
     for z in zones_data:
         if z.get("zone_type") != "content":
             continue
@@ -1892,7 +1894,7 @@ async def _build_grid_core(session_id: str, session: dict) -> dict:
         if not rows:
             continue
 
-        # Extract column_1 (page_ref) cells → zone metadata
+        # Extract column_1 cells that look like page references
         page_refs = []
         page_ref_cell_ids = set()
         for cell in cells:
@@ -1901,6 +1903,8 @@ async def _build_grid_core(session_id: str, session: dict) -> dict:
             text = (cell.get("text") or "").strip()
             if not text:
                 continue
+            if not _PAGE_REF_RE.match(text):
+                continue
             page_refs.append({
                 "row_index": cell.get("row_index"),
                 "text": text,