Better than the best? Answers via model ensemble in density-based clustering