[advancedxy]

What

Add a ProjectNestedArray overload for nested fields without a row-position argument, removing repeated nullptr arguments from recursive projection.

Use ICEBERG_PRECHECK at the record-batch and struct projection entries to validate required row positions. Add one regression test covering nested _pos and _row_id.

Why

Follow-up to the projection review in #1006.

The record-batch check only inspects top-level fields. Metadata fields nested in a struct could therefore reach projection with a null position array. Checking each struct before projection returns an error without adding checks to individual metadata branches.

Testing

• parquet_test passed on the updated main baseline.
• Pre-commit and git diff --check passed.

Limitations and follow-up

The current Iceberg reader supplies the separate position array only to top-level fields. This is an implementation limitation, not a restriction on schema projection. The planned Arrow C++ Parquet reader integration described in #1006 will expose positions in the reader's record batches.

[wgtmac]

Thanks @advancedxy！