feat(ingestion): harden OFF ingestion for bulk seeding

- Add retry/backoff (429 + 5xx, Retry-After aware) to the OFF adapter so
  transient API errors no longer abort a run.
- Clamp bounded text fields (serving_size, net_content_unit,
  country_of_origin) to their column widths in transform; long OFF values
  previously raised StringDataRightTruncation and rolled back the batch.
- Load each record inside a savepoint (load_record_safe) so one malformed
  source record is skipped instead of aborting the whole import; jobs now
  report an errored count.
- Tests for retry behaviour, serving_size clamping, and per-record isolation.

Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>

ingestion/tests/test_load_integration.py

@@ -11,7 +11,7 @@ from pathlib import Path
 
 import pytest
 
-from opengoods.etl.load import default_dsn, ensure_source, load_record
+from opengoods.etl.load import default_dsn, ensure_source, load_record, load_record_safe
 from opengoods.etl.transform import transform
 
 psycopg = pytest.importorskip("psycopg")
@@ -59,3 +59,25 @@ def test_load_record_roundtrip(conn):
     assert prov[0] >= 1
 
     conn.rollback()  # keep the test DB clean
+
+
+def test_load_record_safe_isolates_bad_record(conn):
+    source_id = ensure_source(conn)
+
+    # Unique gtin so the good record is a fresh INSERT, not an upsert/update.
+    good = transform(FIXTURE)
+    good["gtin"] = "4006381333931"
+    assert load_record_safe(conn, good, source_id, FIXTURE) is True
+    after_good = conn.execute("SELECT count(*) FROM product").fetchone()[0]
+
+    # A record whose name violates NOT NULL must not abort the batch.
+    bad = dict(good)
+    bad["gtin"] = "5000112637922"
+    bad["name"] = None
+    assert load_record_safe(conn, bad, source_id, {}) is False
+
+    # The good record survived the bad one's rollback-to-savepoint.
+    after_bad = conn.execute("SELECT count(*) FROM product").fetchone()[0]
+    assert after_bad == after_good
+
+    conn.rollback()  # keep the test DB clean