On the utility of increasing the number of matches in computing sample entropy over short cardiovascular variability series