Incrementing non-matching- but not matching-to-sample is rapidly learned in an automated version of the Odor Span Task