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.

The issue mentions extending OpenMLDB's core to support table aggregation functions in batch mode via Spark dataframes. Start by examining the existing physical plan generation for table aggregation functions in the core codebase. Look for Spark integration points and dataframe handling. Determine what changes are needed in the planner to generate appropriate Spark jobs. Testing will involve verifying the new batch mode works with existing aggregation functions.

Written by the indexing model from the issue text.