Is your feature request related to a problem or challenge?
We want to split IO and CPU. Let's use the ParquetMetaDataPushDecoder as well.
Describe the solution you'd like
Just implement it, so we later can pass it to some IO/morsel queue (do some real prefetching when there is opportunity) and do the conversion later on need.
Describe alternatives you've considered
No response
Additional context
No response