Which advantage justifies the extra work needed to incorporate unit tests into a PySpark application, either through initial job design or refactoring existing code?