Measuring Annotator Agreement Generally across Complex Structured, Multi-object, and Free-text Annotation Tasks