I\'m writing a pandas dataframe to pyarrow.parquet with python2.7, and read it with python3.7. I\'m getting str columns as bytes. If I write the parquet in python3.7 or read