test(e2e): add batch-size cap coverage for SIENTIAPDE-2013

Also fixes the test_activities fixture, which built mongodb_config by
hand and was missing the new required notification_queue_batch_size key
-- every e2e test would have hit a KeyError without this.

New test seeds 5 notifications with a batch size of 2 and confirms three
successive runs return them oldest-first in order (n1,n2 / n3,n4 / n5),
proving the cursor only advances to what was actually returned rather
than skipping documents a capped query didn't fetch. Verified against
real testcontainers Mongo/Redis/Temporal, not mocks.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

e2e/scenarios.md

@@ -110,6 +110,7 @@ No notifications → no SMTP, no Postgres row.
 - Prior timestamp → only newer notifications returned.
 - Query-time filter (SIENTIAPDE-2011) → union of active groups' `levels`, intersection of their `ignore_models_id`; a model ignored by only one of several active groups is not excluded from the fetch.
 - Empty Mongo → no Redis timestamp write.
+- Batch size cap (SIENTIAPDE-2013) → with 5 pending notifications and a batch size of 2, each run returns exactly the next-oldest 2 (sorted ascending by timestamp), and the Redis cursor advances only to what was actually returned, so three successive runs drain the backlog in order instead of skipping anything.
 
 ### ProcessNotifications — `e2e/test_subworkflow_process_notifications.py`