fix: add iceberg_type column and filter all SqlCatalog table operations

[GayathriSrividya]

closes #3337

Summary

Adds the iceberg_type column to the IcebergTables ORM model (also superseding #3263) and extends the type filter to all table operations, matching Java's JdbcCatalog behavior.

Root cause

PR #3263 proposed adding iceberg_type to list_tables, but was not yet merged. Meanwhile #3337 identified that even after that fix, load_table, drop_table, rename_table, and commit_table would still lack the filter. View rows written by iceberg-java or iceberg-rust can therefore bleed into Python's table operations.

Java's JdbcCatalog applies WHERE (iceberg_type = 'TABLE' OR iceberg_type IS NULL) to every table lookup (JdbcUtil.java#L168).

Changes

pyiceberg/catalog/sql.py

• Add iceberg_type: Mapped[str | None] column to IcebergTables model.
• Add ICEBERG_TABLE_TYPE = "TABLE" constant.
• Add idempotent ALTER TABLE iceberg_tables ADD COLUMN iceberg_type migration in _ensure_tables_exist (backward-compatible; catches OperationalError/ProgrammingError if column already exists).
• Set iceberg_type=ICEBERG_TABLE_TYPE on all IcebergTables inserts: create_table, register_table, commit_table (new table path).
• Apply WHERE (iceberg_type = 'TABLE' OR iceberg_type IS NULL) filter to: load_table, drop_table, rename_table, commit_table, list_tables.

tests/catalog/test_sql.py

• Add TestIcebergTypeFilter class with 5 regression tests:
  • test_iceberg_type_set_on_create — verifies new tables get iceberg_type='TABLE'
  • test_list_tables_excludes_view_rows
  • test_load_table_ignores_view_rows
  • test_drop_table_ignores_view_rows
  • test_rename_table_ignores_view_rows

Validation

• make lint ✓
• pytest tests/catalog/test_sql.py → 17 passed ✓

[ebyhr]

I don't think we should add a new column in _ensure_tables_exist method. It is counterintuitive. The initialization logic in #3263 looks better than this PR.