Adaptive bias for dissensus in nonlinear opinion dynamics with application to evolutionary division of labor games