improve shape tests for 2plus1D

[rtimms]

Sometimes shape test gives an error because the domain of a variable isn't set to current collector in 2D, when it should be. In practice the object ends up being the correct shape, just the domain it is labelled with is wrong. e.g. in inverse_butler_volmer the line eta_r = (2 / ne) * pybamm.Function(np.arcsinh, j / (2 * j0)) gives a shape error. j has the expected shape, but j0 doesn't as it contains a broadcast whose child had domain [] but should be ["current collector"]

[valentinsulzer]

I've fixed this in #546 (I think)