fix(chunker): תיקון זיהוי כותרות טענות הצדדים ביחיד/נקבה + שלוש שכבות הגנה

**סיבת-שורש:** רג'קס respondent_claims כיסה רק צורת רבים (המשיבים/המשיבין),
ולא יחיד נקבה (המשיבה) ויחיד זכר (המשיב). הכותרת "טענות המשיבה:" בתיק 8181-21
נבלעה לתוך מקטע ruling → חולצה כהלכה שגויה.

**שלוש שכבות הגנה:**
1. chunker.py — הרחבת SECTION_PATTERNS לכסות יחיד/זכר/נקבה + תגובת/תשובת
2. halacha_extractor.py — עיגון חיובי: drop ל-ruling chunks שלפני legal_analysis
3. halacha_quality.py — FLAG_PARTY_CLAIM: זיהוי שפת-טענות-צד בציטוט התומך

**היקף:** 93 תיקים עם 0 chunks של טענות (כנראה בגלל כותרות ביחיד שלא הוכרו);
628 הלכות מאושרות מתיקים אלה — חלקן עשויות להיות תקינות, יש לעשות re-chunk.

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

mcp-server/src/legal_mcp/services/halacha_extractor.py

@@ -638,12 +638,20 @@ async def _select_extractable_chunks(
     extraction. Previously the fallback took *all* chunks, re-admitting exactly
     the sections the primary filter excludes.
 
+    Positive-anchor guard: when a document has a "דיון/הכרעה" section
+    (section_type='legal_analysis'), extraction starts from that section
+    onwards.  Any 'ruling' chunks that appear BEFORE the first legal_analysis
+    chunk by position are dropped — they most likely result from a party-claims
+    section whose header was not recognised by the chunker and was therefore
+    absorbed into the preceding section.
+
     Returns ``(chunks, used_fallback)`` so the caller can log the fallback once.
     """
     chunks = await db.list_precedent_chunks(
         case_law_id, section_types=EXTRACTABLE_SECTIONS,
     )
     if chunks:
+        chunks = _apply_discussion_anchor(chunks)
         return chunks, False
     all_chunks = await db.list_precedent_chunks(case_law_id)
     filtered = [
@@ -653,6 +661,37 @@ async def _select_extractable_chunks(
     return filtered, True
 
 
+def _apply_discussion_anchor(chunks: list[dict]) -> list[dict]:
+    """Drop 'ruling' chunks that precede the first 'legal_analysis' chunk.
+
+    In Israeli planning-committee decisions the discussion section
+    (דיון / הכרעה / דיון והכרעה) always comes after the parties' claims.
+    A 'ruling'-labelled chunk that appears before the discussion is a strong
+    signal that a party-claims section was silently absorbed into it (chunker
+    regex didn't match the header).  Dropping those early 'ruling' chunks is
+    safe because all reasoning content falls at or after the 'דיון' anchor.
+    """
+    analysis_indices = [
+        c["chunk_index"] for c in chunks
+        if c.get("section_type") == "legal_analysis"
+    ]
+    if not analysis_indices:
+        return chunks
+    first_analysis = min(analysis_indices)
+    filtered = [
+        c for c in chunks
+        if not (c.get("section_type") == "ruling" and c["chunk_index"] < first_analysis)
+    ]
+    dropped = len(chunks) - len(filtered)
+    if dropped:
+        logger.info(
+            "halacha_extractor: positive-anchor guard dropped %d pre-discussion "
+            "'ruling' chunk(s) (first legal_analysis at chunk_index=%d)",
+            dropped, first_analysis,
+        )
+    return filtered
+
+
 async def _extract_impl(case_law_id: UUID, force: bool = False,
                         effort: str | None = None) -> dict:
     """Core extraction (caller holds the global advisory lock for the duration).