SqlCatalog table operations should filter on iceberg_type #3337

Feature Request / Improvement

In #3263, the iceberg_type column was added to the SQLCatalog and filters it in list_tables, but the other table operations (i.e. load_table, drop_table, rename_table, commit_table) don't check it yet. That means a view row written by iceberg java or iceberg-rust can bleed in pythons SQLCatalog table operations.

Java's V1 catalog applies WHERE (iceberg_type = 'TABLE' OR iceberg_type IS NULL) everywhere (JdbcUtil.java#L168). We should do the same.

Probably worth extracting a small helper since the predicate would now appear in ~5+ places.

Repro steps:

catalog = SqlCatalog("test", uri="sqlite:///...", warehouse="...")
catalog.create_namespace("ns")

# bypass and sim java/rust writing a view row
with catalog.engine.connect() as conn:
    conn.execute(text(
        "INSERT INTO iceberg_tables VALUES "
        "('test', 'ns', 'a_view', 's3://fake/m.json', NULL, 'VIEW')"
    ))
    conn.commit()

catalog.drop_table(("ns", "a_view"))  # Silently deletes the view row :/