AsyncParquetReader Pre-Fetch Next Row Group

[tustvold]

Is your feature request related to a problem or challenge? Please describe what you are trying to do.

The AsyncParquetReader currently fetches bytes and then performs decode, in scenarios where scanning an entire parquet file it may be beneficial to fetch data for the next row group in parallel to decoding the current row group.

Describe the solution you'd like

Describe alternatives you've considered

Additional context

One of the experiments on https://github.com/apache/arrow-rs/pull/1509/files#diff-f6b1a106d47a16504d4a16d57a6632872ddf596f337ac0640a13523dccc2d4d4R448 implemented something along these lines using a peekable stream

[tustvold]

I think this is better handled at the query engine level, which is better placed to achieve the desired parallelism by processing the row groups in separate partitions where appropriate