Amplitude -> Power spectrum

@reimar, @pfrank, @kjako, @mtr how do you like this change? The most important change is this one: 43761bb6

It boils down to the fact that we model the actual power spectrum and not the amplitude with the integrated Wiener process. This saves one nonlinearity in the forward model and simplifies the interface, i.e. people do not need to remember to divide the values for loglogavgslope by two.