Fix calculation of L2 error in angeli test

In postTimeStep the L2 error is calculated using a difference to the analytical solution. Without this commit, the L2 error is calculated using a difference between the numerical solution at a certain time and the analytical solution one time step ahead. With this commit, the numerical and analytical solution used to calculate the L2 error should correspond to the same time. (This was spotted by @AndreaVescovini)

That this error came up in the first place is probably related to the somewhat confusing implementation of the L2 error calculation and analytical solution creation and handling of its time dependency (see also #646 (closed))