My real analysis is extremely rusty, but could you use Urysohn's lemma to get an arbitrary continuous "step" function and then integrate that.