Support table aggregation functions for Batch mode

Now OpenMLDB core support generate physical plan for table aggregation functions. We should add support for batch mode by generate planner to run table aggregation functions over Spark dataframes.