Question on pig loader read parquet file

When I use spark save parquet file, schema like this

```

optional group attref (LIST) {
repeated group list {
optional group element {
optional binary nid (UTF8);
optional binary nss (UTF8);
}
}
}
```

And then use parquet-pig-bundle to read this file, the read function can work, but when i need to access "nid" it have some problem

If I read other file save by pig-storer, and need nid list, pig command is:
```

B = foreach A generate value.addr.clientIp_bag.clientIp, value.guid , value.attref.nid;
```
but read spark save version I need use this:
```

B = foreach M generate value.addr.clientIp, value.guid , flatten(value.attref);
C = foreach B generate clientIp, guid, attref::element.nid;
```
and this command will flatten column

My question is pig loader have some problem when loading parquet file(save by spark)

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

_{**Note**: *This issue was originally created as [PARQUET-1172](https://issues.apache.org/jira/browse/PARQUET-1172). Please see the [migration documentation](https://issues.apache.org/jira/browse/PARQUET-2502) for further details.*}

报告提到了 parquet-pig-bundle，并将 Spark 写入的 LIST schema 与 Pig-storer 的输出进行对比；先复现这两种读取方式，并跟踪 loader 如何暴露 attref 和 nid。完成标准是：Spark 生成的文件能够在不进行不必要的扁平化的情况下，支持预期的 nid 访问，并且涵盖所报告的 schema 和命令。

由索引模型根据 Issue 内容生成。