fix: handle upsert after schema evolution with added columns

[GayathriSrividya]

closes #3105

Summary

• Fix upsert row-matching when source schema has newly-added non-key columns after schema evolution.
• Avoid casting the entire source table to the target schema in upsert row comparison.
• Add a regression test for upsert after add_column evolution.

Root cause

get_rows_to_update cast the full source table to the matched target batch schema before joining. When source contains a newly-added column, target scan batches may not include that column, causing schema-name mismatch errors.

Changes

• In pyiceberg/table/upsert_util.py:
  • Cast only join columns to target join-column schema.
  • Keep join key order stable by using join_cols directly.
  • Treat missing non-key columns in target rows as None during comparison.
• In tests/table/test_upsert.py:
  • Add test_upsert_after_schema_add_column regression test.

Validation

• make lint
• /path/.venv/bin/python -m build
• /path/.venv/bin/python -m pytest tests/table/test_upsert.py -xvs