No need to read full RowGroup when using RowRanges

### Describe the enhancement requested

I'm using ParquetFileReader and would like to read the few first line of a parquet file to show a preview to the user.
I'm doing this :
PageReadStore rowGroup = reader.readFilteredRowGroup(0, RowRanges.createSingle(10));
and then :
ColumnReadStore colReadStore = new ColumnReadStoreImpl(rowGroup,
new GroupRecordConverter(schema).getRootConverter(),
schemaLecture, ...);
etc.

The problem is when calling readFilteredRowGroup(), it reads the full row group in memory, which in my case is slow because the parquet file is on a network drive and has quite big row group. On top of that, it consumes RAM for nothing.

Is this an issue or could it be improved ?

Thanks in advance.
Olivier

### Component(s)

_No response_

从 ParquetFileReader.readFilteredRowGroup() 开始，跟踪加载 row group 时如何处理 RowRanges.createSingle(10)。确定在哪里读取了完整的 row group，以及范围过滤如何能够限制网络读取和内存使用；当无需加载完整的 row group 即可读取请求的行时，就算完成。

由索引模型根据 Issue 内容生成。