How many copies are enough? Using citation studies to limit journal holdings