Comment on “Symplectic integration of magnetic systems”: A proof that the Boris algorithm is not variational