Fix bug where we failed to delete old push actions (matrix-org#13194)

This happened if we encountered a stream ordering in `event_push_actions` that had more rows than the batch size of the delete, as If we don't delete any rows in an iteration then the next time round we get the exact same stream ordering and get stuck.

Author: erikjohnston

changelog.d/13194.bugfix

@@ -0,0 +1 @@
+Fix bug where rows were not deleted from `event_push_actions` table on large servers. Introduced in v1.62.0.

synapse/storage/databases/main/event_push_actions.py

@@ -1114,7 +1114,7 @@ def remove_old_push_actions_that_have_rotated_txn(
             txn.execute(
                 """
                 SELECT stream_ordering FROM event_push_actions
-                WHERE stream_ordering < ? AND highlight = 0
+                WHERE stream_ordering <= ? AND highlight = 0
                 ORDER BY stream_ordering ASC LIMIT 1 OFFSET ?
             """,
                 (
@@ -1129,10 +1129,12 @@ def remove_old_push_actions_that_have_rotated_txn(
             else:
                 stream_ordering = max_stream_ordering_to_delete
 
+            # We need to use a inclusive bound here to handle the case where a
+            # single stream ordering has more than `batch_size` rows.
             txn.execute(
                 """
                 DELETE FROM event_push_actions
-                WHERE stream_ordering < ? AND highlight = 0
+                WHERE stream_ordering <= ? AND highlight = 0
                 """,
                 (stream_ordering,),
             )