Spark split one dataframe into multiple. If not provided, default limit value is -1