Expose row group index in Parquet readers

[uros7251brick]

Describe the enhancement requested

Description

Similar to how getCurrentRowIndex() was introduced to expose the current row's file-level index, this adds getCurrentRowGroupIndex() to expose the index of the row group currently being read.

New API

• ParquetFileReader.getCurrentRowGroupIndex() — returns the 0-based index of the last row group read via readNextRowGroup() / readNextFilteredRowGroup(). Returns -1 before any row group has been read.
• ParquetReader.getCurrentRowGroupIndex() — same semantics, for the high-level record reader.
• ParquetRecordReader.getCurrentRowGroupIndex() — same, for the Hadoop MapReduce record reader.

The returned index is the actual file-level row group index, meaning it correctly reflects gaps when empty row groups are skipped (e.g. if row group 1 is empty, the indices reported will be 0, 2, ... not 0, 1, ...).

Motivation

Engines like Apache Spark need to know which row group a record belongs to — for example, to expose row group metadata as a hidden column, or to correlate records with row group-level statistics. Without this API, callers have no way to determine the current row group index during sequential reads.

Component(s)

No response