A note on the optimality of decomposable entanglement witnesses and completely entangled subspaces