fix: poll for fault-manager operations in turtlebot3 smoke test

In Gazebo-heavy demos, the fault_manager ROS 2 services take longer
to be discovered by runtime graph introspection. Use poll_until with
a 30s timeout instead of an instant assert_non_empty_items check.

Author: bburda

tests/smoke_test_turtlebot3.sh

@@ -52,7 +52,13 @@ assert_non_empty_items "/apps/medkit-gateway/data"
 
 section "Operations"
 
-assert_non_empty_items "/apps/medkit-fault-manager/operations"
+# fault_manager services may take extra time to be discovered in Gazebo-heavy demos
+echo "  Waiting for fault-manager operations to appear (max 30s)..."
+if poll_until "/apps/medkit-fault-manager/operations" '.items | length > 0' 30; then
+    pass "GET /apps/medkit-fault-manager/operations returns non-empty items"
+else
+    fail "GET /apps/medkit-fault-manager/operations returns non-empty items" "items still empty after 30s"
+fi
 
 section "Configurations"