Add pub method to the BatchedParquetReader to read row group directly

[deanm0000]

Description

I often save big parquet files with BatchedWriter forcing each row group to contain only a single node_id. When reading those files I use the ParquetAsyncReader to get the row group metadata so that I know which node_ids are in the file. For reference I do this

    let mut async_reader = ParquetAsyncReader::from_uri(cloud_path,None,None).await.unwrap();
    let metadata =async_reader.get_metadata().await.unwrap();
    let meta_cloned = Arc::clone(&metadata);
    let row_groups = &meta_cloned.row_groups;
    let mut nodes:Vec<u64> = vec![];
    row_groups.into_iter().for_each(|(i, rg)| {
        let columns = rg.columns_under_root_iter("node_id").unwrap();
        columns.into_iter().for_each(|col| {
            let c=col.metadata();
            let stats=c.statistics.clone().unwrap();
            let min=stats.min_value.clone().unwrap();
            let array: [u8; 8] = min.try_into().expect("Vec<u8> is not 8 bytes long.");
            let node=u64::from_le_bytes(array);
            nodes.push(node);
        });
    });

As it is now I can't (or don't see how) to use the same async_reader to read the data from each row group. Instead I have to do a LazyFrame::scan_parquet with a filter.

The async_reader already has a RowGroupFetcher which has fetch_row_groups but they're not public so can't be used directly.

This would (I think) also alleviate the need to use bigidx since none of my row groups are that big.