fix: preserve bigint batch identities in news AI queries

Use BIGINT casts for JSON batch references in queue filters, retries and legacy job reconciliation. Cover real large batch references and absent 64-bit IDs in isolated PostgreSQL acceptance.

backend/src/alphapulse/news_risk/reprocess.py

@@ -1,5 +1,5 @@
 """Explicit bounded historical jobs, sharing the normal news pipeline."""
-from sqlalchemy import select, text, update
+from sqlalchemy import select, text, update, BigInteger
 from alphapulse.core.db import SessionLocal
 from alphapulse.models import NewsItem
 from alphapulse.models.risk import RiskSettings, ReprocessJob, ReprocessItem, ArticleAnalysis, NewsSummary, EventMember
@@ -68,7 +68,7 @@ async def process_job(item_index=None):
                 inputs = await reading_inputs(session, datetime.now(UTC)-timedelta(hours=72))
                 key = briefing.fingerprint(briefing.semantic_packages(inputs), job.settings['configurations']['outlook']['rules'])
                 snapshot = await session.scalar(select(RiskSnapshot).where(
-                    RiskSnapshot.quotes['job_id'].as_integer() == job.id,
+                    RiskSnapshot.quotes['job_id'].astext.cast(BigInteger) == job.id,
                     RiskSnapshot.quotes['fingerprint'].as_string() == key).order_by(RiskSnapshot.id.desc()).limit(1))
                 if snapshot is None or snapshot.state == 'processing' or snapshot.quotes.get('retry_requested'): return
                 if snapshot.state == 'failed':