ParquetWriter: opaque error when record batch schema is missing field ids #2654

What's the problem?

When matching an Arrow record batch to an Iceberg schema by field id, ParquetWriter (via ArrowArrayAccessor::field_partner in crates/iceberg/src/arrow/value.rs) reads the PARQUET:field_id (PARQUET_FIELD_ID_META_KEY) metadata off each Arrow field. When a caller hands it a record batch whose Arrow schema was built by hand--without that metadata--the match fails with:

DataInvalid => Field id N not found in struct array

The message points at the symptom, not the cause: the Arrow field carries no field id at all. The right way to build the batch schema is to derive it from the table schema via table.metadata().current_schema().as_ref().try_into(), which stamps the field-id metadata--but nothing surfaces that, so it's an easy trap for downstream consumers.

Reported by @malon64 while testing #2185 from a downstream Rust ingestion tool.

Proposed fix

Enrich the error at its source in ArrowArrayAccessor::field_partner. In FieldMatchMode::Id, name the field and note the likely cause (a missing PARQUET:field_id), pointing at current_schema().as_ref().try_into(); FieldMatchMode::Name gets a matching by-name message.

Message-only change--the error fires in exactly the same cases as before, so there is no behavior change. (An earlier attempt validated the batch schema up front in the writer, but that wrongly rejected valid batches carrying extra non-Iceberg columns, e.g. the datafusion _partition column, so the source-error approach is the correct one.)

Willingness to contribute

I can contribute this fix independently--PR incoming.