write_dataframe_async with increment=True should aggregate before splitting dataframe #1307
Description
When using write_dataframe_async with increment=True, the aggregation/incrementation should happen before the dataframe is split into smaller chunks for parallel processing.
Currently, the aggregation appears to happen per chunk, meaning that if the original dataframe contains duplicate records (same intersection), these duplicates might be distributed across different chunks. As a result, parallel writes can overwrite each other instead of properly incrementing values.