The Eclipse Score, or, how to determine validators who are not effectively representing their delegators via voting