Tiny optimization in find()

The PyMongo Cursor returns results from a batch by doing self.__data.pop(0) for each element. This requires quadratic time over the size of the batch. In tests, Python can do pop(0) on a 16k-element list in 3 or 4 microseconds, so it's not a big problem. But reversing the list and doing pop() might save a couple microseconds for big queries.