Core: Improve error for null/unknown schema types in table creation

[010Soham]

What changed

• Add a client-side validation error when a PyArrow null type is used with format version < 3, including the field path and a hint to use a concrete type or format-version 3.
• Add a unit test to verify the clear error message when converting a PyArrow schema with a null field.

Why

• Prevents misleading REST errors like “Cannot parse type string… unknown” and points directly to the offending field.

Testing

• make lint
• uv run pytest tests/catalog/test_base.py -k rejects_null_type -v
• make test (timed out at ~42%)

Closes #2539

[kevinjqliu]

LGTM! Thanks pa.null() has been a long standing issue. This should provide some clarity

[kevinjqliu]

this is the full error message

'Null type (pa.null()) is not supported in Iceberg format version 2. Field: n1. Use a concrete type (string, int, boolean, etc.) or set format-version=3.'

[geruh]

nit: wdyt about future proofing with a message like: Requires v3+ or use a concrete type.