fix(tests): the column guard names join tables as _BACKED_UP holds them (#3182)

_BACKED_UP carries REAL table names — project_rulebook_subscriptions,
project_rule_suppressions, project_topic_suppressions,
project_rulebook_exclusions — not the shorter keys the payload uses for the
same sections. The registry-coverage assertion used the payload spelling and
reported four tables unguarded.

The integration round trip passed on this run, which is the half that matters:
the real restore_full_backup remaps arose_from_id onto the restored origin.

tests/test_services_backup.py

@@ -242,9 +242,13 @@ def test_the_column_guard_covers_every_table_with_a_row_helper():
     with a new helper, and not to the registry, would be unguarded and look
     guarded. Join tables have no model class and carry both their columns by
     construction, so they are the only permitted absences."""
+    # REAL table names, as _BACKED_UP holds them — not the shorter keys the
+    # payload uses for the same sections. Getting this wrong is what the guard
+    # caught on its own first run.
     join_tables = {
-        "rulebook_subscriptions", "rule_suppressions",
-        "topic_suppressions", "rulebook_exclusions", "rule_systems",
+        "project_rulebook_subscriptions", "project_rule_suppressions",
+        "project_topic_suppressions", "project_rulebook_exclusions",
+        "rule_systems",
     }
     covered = set(_column_guard_targets()) | join_tables
     assert set(backup._BACKED_UP) - covered == set()