Hi, I am currently using pyarrow to store pandas datasets. Due to the very large amount of data it would be useful to use a lower precision for floats. I saw that it is implemented in the parquet s...