Use non-zero weight limit in benchmarking `pallet_xcm::execute()` so the given XCM actually executes.