Conforms pandas to "correct" datatypes to ensure data in/out using CSV, JSONL and Parquet is read the same (using arrow). - View it on GitHub