[python] Introduce 'nested_update' aggregator function

[PyRSA]

Purpose

Introduce the nested_update field aggregator for pypaimon.

This change adds support for the nested_update aggregation function in the Python reader, aligning its behavior with the Java implementation. It supports nested-key based updates, optional sequence-field based conflict resolution, configurable nested-key null handling, and count limiting.

Dependency

This PR depends on #8561 and reuses several shared utility methods introduced there. Please review and merge #8561 first.

Tests

• python -m pytest pypaimon/tests/test_field_aggregators.py::FieldNestedUpdateAggTest -q
• git diff --check

[JingsongLi]

• RowType.get_field_index() returns DataField.id, but the ProjectedRow of nested key/sequence uses it as an array index.
• The actual nested field ID is a global ID and does not guarantee continuity from 0.
• Reproducing locally using field IDs 10, 20: Both generate mapping=[10], then directly throw an IndexError: Position 10 is out of bounds for row arity 2.
• The existing tests use 0, 1, 2, which obscures the issue. The ordinal field should be returned, and non-sequential ID tests should be added.

[PyRSA]

• RowType.get_field_index() returns DataField.id, but the ProjectedRow of nested key/sequence uses it as an array index.
• The actual nested field ID is a global ID and does not guarantee continuity from 0.
• Reproducing locally using field IDs 10, 20: Both generate mapping=[10], then directly throw an IndexError: Position 10 is out of bounds for row arity 2.
• The existing tests use 0, 1, 2, which obscures the issue. The ordinal field should be returned, and non-sequential ID tests should be added.

Good catch, thanks! I've fixed RowType.get_field_index() to return the ordinal field position instead of DataField.id, and added a regression test using non-sequential field IDs to cover this case.

[JingsongLi]

nested_update assumes that the elements of ARRAY<ROW> are InternalRow; however, the PyArrow→Polars reader actually outputs a list[dict], and calling it directly results in an AttributeError: ‘dict’ object has no attribute ‘get_field’.

[PyRSA]

nested_update assumes that the elements of ARRAY<ROW> are InternalRow; however, the PyArrow→Polars reader actually outputs a list[dict], and calling it directly results in an AttributeError: ‘dict’ object has no attribute ‘get_field’.

Thanks for pointing this out.
I updated the implementation to support both InternalRow and dict representations of nested rows.

Instead of accessing fields directly through InternalRow.get_field(), the aggregator now uses a common FieldProjection helper to extract configured fields from either row representation. This keeps the aggregation logic unchanged while making it compatible with the PyArrow → Polars read path.

[JingsongLi]

+1

[JingsongLi]

Please resolve conflicts.

[PyRSA]

Please resolve conflicts.

Resolved the merge conflicts and updated the branch. Thanks!