Two Nil Python rows asserted behaviour that was later changed on purpose

Found 2026-10-19 in the full row sweep. Neither row could fail while the recipe was stopping at row 593, so both went stale without anyone seeing it.

• test_nilpy_optional_int_none expected int(None) to print 0. CPython raises TypeError, and 0f6416c42f made pxx raise too. The row now prints the None value and catches the TypeError (7, 42, None, TypeError, matching CPython). pyvar_to_int's internal None→0 sentinel is unchanged.
• test_nilpy_dataclass_frozen_fail expected @dataclass(frozen=True) to be refused. ae615953df (10-09) deliberately changed that to accepted-with-warning for the TSP port. The row now compiles the file and greps for the warning.