Judging the conversation in `turns` (with the conversation-level fields in `test_case`) against the `evaluation_steps`, which rubric level best describes it?