When dealing with large amounts of profile data, json and csv has limits.

I have started implementing output_parquet. I'm not a very experienced with C++ but I'm making progress.

Would this be a relevant feature for this project?