use apache arrow for feather/parquet instead of feather directly

[sebkopf]

https://github.com/apache/arrow/tree/main/r
https://github.com/wesm/feather

see the blog post about move of feather to arrow here:
https://arrow.apache.org/blog/2019/08/08/r-package-on-cran/

generally --> consider switching to feather based iso_save() to automatically provide R and python support (not sure exactly how that would work but will look into it)

Arrow framework:

• R: https://arrow.apache.org/cookbook/r/
• python: https://arrow.apache.org/docs/python/

note that polars (see #186) is already build on arrow framework

big question is whether to export to feather (faster read/write) or parquet (smaller file storage):

• https://www.r-bloggers.com/2022/05/comparing-performances-of-csv-to-rds-parquet-and-feather-file-formats-in-r/
• https://www.jumpingrivers.com/blog/parquet-file-format-big-data-r/