Computing Possible and Necessary Winners from Incomplete Partially-Ordered Preferences