My understanding is that it cannot. You could fake it somewhat but aggregating all level one items to level two and then running a multigroup model in which group one is the CFA at level 1 and group 2 is the CFA at level 2. My colleagues had a bit of a back and forth about this idea with the creators of mplus a little while back and this was the compromise was arrived at. I have my concerns as i think the SEs in such a model will be too small at L1 and I think the same will apply to the chi-square test.