Relevant references:
Todo, practically: in read_dataframe, when
use_arrow is False
- GDAL > 3.12.1
- for files using the parquet or arrow driver
Use the new open option LISTS_AS_STRING_JSON=YES. When using this open option, list columns will be returned by GDAL as JSON, which are converted to lists again in read_dataframe, but then None values in those list will be correct.