Fix xfailing regions test

[astrofrog]

In 486e113 I had to xfail a test that is failing because of changes in the regions translation.

@jsub1 - would you be able to take a look at this?

[jsub1]

@astrofrog The translation is supposed to fail because one of the ROIs is defined on the main component. 7b79fc1 removed where we checked for that. Does it make sense to try and expand translation to astropy regions for non-2D datasets?

[astrofrog]

@jsub1 - for some reason the translation wasn't failing, which is why I had to xfail the test. I think ideally we'd like to get the translation to work for 3D+ datasets, I'm not sure though how we convey the information about which axes are covered by the regions. Feel free to take a stab at this!