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/opengoods/adapters/openfoodfacts.py

@@ -27,6 +27,9 @@ _DEFAULT_MIN_INTERVAL = 4.0
 _API_URL = "https://world.openfoodfacts.org/api/v2/product/{barcode}.json"
 _SEARCH_URL = "https://world.openfoodfacts.org/api/v2/search"
 
+# HTTP statuses worth retrying: rate limiting and transient server errors.
+_RETRY_STATUS = frozenset({429, 500, 502, 503, 504})
+
 # Fields requested from the search API so a returned product can be transformed
 # without an extra per-barcode round trip.
 _SEARCH_FIELDS = (
@@ -46,9 +49,13 @@ class OpenFoodFactsAdapter:
         self,
         client: httpx.Client | None = None,
         min_interval: float = _DEFAULT_MIN_INTERVAL,
+        max_retries: int = 4,
+        backoff_base: float = 2.0,
     ) -> None:
         self._client = client or httpx.Client(headers={"User-Agent": USER_AGENT}, timeout=30.0)
         self._min_interval = min_interval
+        self._max_retries = max_retries
+        self._backoff_base = backoff_base
         self._last_call = 0.0
 
     def _throttle(self) -> None:
@@ -58,11 +65,49 @@ class OpenFoodFactsAdapter:
             time.sleep(wait)
         self._last_call = time.monotonic()
 
+    def _get(self, url: str, params: dict | None = None) -> httpx.Response:
+        """GET with throttling and retry/backoff on transient errors.
+
+        Retries on connection/timeout errors and on retryable HTTP statuses
+        (429 and 5xx, which OFF returns intermittently when overloaded), using
+        exponential backoff that honours a ``Retry-After`` header when present.
+        """
+        last_exc: Exception | None = None
+        for attempt in range(self._max_retries + 1):
+            self._throttle()
+            try:
+                resp = self._client.get(url, params=params)
+            except httpx.TransportError as exc:
+                last_exc = exc
+            else:
+                if resp.status_code < 400 or resp.status_code not in _RETRY_STATUS:
+                    resp.raise_for_status()
+                    return resp
+                last_exc = httpx.HTTPStatusError(
+                    f"retryable status {resp.status_code}", request=resp.request, response=resp
+                )
+            if attempt < self._max_retries:
+                retry_after = self._retry_after(last_exc)
+                time.sleep(retry_after if retry_after is not None else self._backoff_base**attempt)
+        assert last_exc is not None
+        raise last_exc
+
+    @staticmethod
+    def _retry_after(exc: Exception | None) -> float | None:
+        resp = getattr(exc, "response", None)
+        if resp is None:
+            return None
+        value = resp.headers.get("Retry-After")
+        if not value:
+            return None
+        try:
+            return float(value)
+        except ValueError:
+            return None
+
     def fetch_barcode(self, barcode: str) -> dict | None:
         """Fetch a single product by barcode; return the raw `product` dict."""
-        self._throttle()
-        resp = self._client.get(_API_URL.format(barcode=barcode))
-        resp.raise_for_status()
+        resp = self._get(_API_URL.format(barcode=barcode))
         payload = resp.json()
         if payload.get("status") != 1:
             return None
@@ -91,8 +136,7 @@ class OpenFoodFactsAdapter:
         ``last_modified_t`` they processed as the next watermark.
         """
         for page in range(1, max_pages + 1):
-            self._throttle()
-            resp = self._client.get(
+            resp = self._get(
                 _SEARCH_URL,
                 params={
                     "fields": _SEARCH_FIELDS,
@@ -101,7 +145,6 @@ class OpenFoodFactsAdapter:
                     "page_size": page_size,
                 },
             )
-            resp.raise_for_status()
             products = resp.json().get("products") or []
             if not products:
                 return