Add option to import by row group (not just chunk size)

[olivialynn]

Suggested in the working group meeting today

[hombit]

A bit more context: allow iterating input parquet files by row-groups instead of (along with?) number or rows. We may add an additional argument to ParquetReader and ParquetPyarrowReader for that.

I think it can be also useful for nested data sets, for example for HATS re-importing