Pool SeekableInputStreams in ParquetFileReader

If https://issues.apache.org/jira/browse/PARQUET-1982 goes through, then we could allow parallel reading of row groups with a pool of SeekableInputStreams. This would significantly boost performance for applications that read data at random positions from a large file.

I've already developed a patch that would enable this functionality. I will link the merge request in the next few days.

Is there a related ticket that i have overlooked?

**Reporter**: [Felix Schmalzel](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=fschmalzel) / @fschmalzel

_{**Note**: *This issue was originally created as [PARQUET-1983](https://issues.apache.org/jira/browse/PARQUET-1983). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

首先查看 PARQUET-1982 和现有的 ParquetFileReader 实现，以了解前置条件和当前的流处理方式。通过 SeekableInputStreams 池启用并行读取 row group，并验证随机位置读取的性能表现，即视为完成。

由索引模型根据 Issue 内容生成。