fix: Don't use _position_to_field_name (#3917)

Don't use internal fields :)

I want to remove this one in
apache/iceberg-python#1768. It should be okay
since the `Record` also has a `__len__`.

Author: Fokko

daft/iceberg/iceberg_scan.py

@@ -129,7 +129,7 @@ def _iceberg_record_to_partition_spec(
     ) -> daft.recordbatch.RecordBatch | None:
         partition_fields = iceberg_partition_spec_to_fields(self._table.schema(), spec)
         arrays = dict()
-        assert len(record._position_to_field_name) == len(partition_fields)
+        assert len(record) == len(partition_fields)
         for idx, pfield in enumerate(partition_fields):
             field = Field._from_pyfield(pfield.field)
             field_name = field.name

pyproject.toml

@@ -29,7 +29,7 @@ azure = []
 deltalake = ["deltalake"]
 gcp = []
 hudi = ["pyarrow >= 8.0.0"]
-iceberg = ["pyiceberg >= 0.4.0", "packaging"]
+iceberg = ["pyiceberg >= 0.7.0", "packaging"]
 lance = ["lancedb"]
 numpy = ["numpy"]
 pandas = ["pandas"]