[BUG] Use iceberg FileIO to access parquet files.

[liurenjie1024]

Currently in iceberg's parquet reader, we are still using hadoop fs to access parquet file. This has several limitations:

1. It can't support as much storage systems as iceberg does. Though there are some workaround for s3, there are more underlying storages that iceberg supports.
2. It doesn't contains necessary information such credentails as in FileIO

We should use iceberg's FileIO to access underlying parquet files.