arrow

GH-44668: [Docs] Fix ColumnChunkMetaData offset documentation in pyarrow (#44670)

### Rationale for this change

The pyarrow documentation of ColumnMetaData is contradicting the C++ implementation.

The pyarrow [documentation](https://arrow.apache.org/docs/dev/python/generated/pyarrow.parquet.ColumnChunkMetaData.html#pyarrow.parquet.ColumnChunkMetaData.data_page_offset) says:
The data_page_offset and dictionary_page_offset are relative to the column chunk offset

The C++ [comments in the code](https://github.com/apache/arrow/blob/df24a8225999896eb03db280354fbff42dfea0f5/cpp/src/generated/parquet_types.h#L2896) say:
The offsets are byte offsets from the beginning of the file to first data_page / dictionary_page

### What changes are included in this PR?

Update comments that `data_page_offset` and `dictionary_page_offset` are relative to start of file

### Are these changes tested?

Verified locally that C++ code comments are correct

### Are there any user-facing changes?

Documentation

GitHub Issue: https://github.com/apache/arrow/issues/44668
* GitHub Issue: #44668

Authored-by: Georg Kreuzmayr <g.kreuzmayr@gmail.com>
Signed-off-by: mwish <maplewish117@gmail.com>