Spark union two dataframes with different columns. Anything better solution ...