Support splitting parquet input files

[richox]

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

A big parquet input file should be splitted into smaller partitions, which is not currently supported.

Describe the solution you'd like

Implement ParquetExec with filename, offset and length parameters.

Describe alternatives you've considered

does parquet2 provide better supports?

Additional context

[yjshen]

DataFusion imposes the current restriction. There is an umbrella ticket containing this problem: apache/datafusion#944. You can fix this in DataFusion.

[yjshen]

"Enable file block granularity processing (row groups or offset range processing instead of currently per-file bases)."

[yjshen]

Trying to fix in upstream arrow-rs and arrow-datafusion.
arrow-rs: apache/arrow-rs#1389