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

The current ParquetRecordBatchReader::try_new_with_row_groups function always uses ProjectionMask::all(), which forces reading all columns from the parquet file even when only specific columns are needed. This is inefficient for use cases where you only want to read a subset of columns, as it results in unnecessary I/O and memory usage.

Describe the solution you'd like

Modify the try_new_with_row_groups function to accept a projection: &ProjectionMask parameter instead of hardcoding ProjectionMask::all(). This will allow callers to specify exactly which columns they want to read, enabling more efficient data access patterns.

Describe alternatives you've considered

Additional context