Comment on "Role of vortex strings in the three-dimensional O(2) model