Rewrite ParquetRecordBatchReader (sync api) in terms of the PushDecoder #8678
Description
Is your feature request related to a problem or challenge? Please describe what you are trying to do.
We added the ParquetPushDecoder in #7997
One of the rationales is to avoid duplicating the control logic between the Async Reader and the Sync Reader.
However, it actually (temporarily) makes the problem worse by adding a 3rd copy of the control logic ala the xkcd standards effect
Once we have completed this ticket and the following one, there will be one control loop:
Describe the solution you'd like
Rewrite ParquetRecordBatchReader in using the ParquetPushDecoder
Describe alternatives you've considered
The IO pattern in the sync decoder is different than the async decoder -- it evaluates predicates on all row groups first and then decodes each row group so the push decoder will take some finagling
Additional context