Sample size calculation for studies comparing binary outcomes using historical controls