Saturate (np.clip) unstructured controller?

[ryancoe]

In reviewing #273, I noticed that we could actually saturate controller_unstructured the same way @cmichelenstrofer did for the PID controllers using np.clip. This would, in theory, produce the same result as a constraint applied to the optimization solution. However (as pointed out by @cmichelenstrofer) you would sort of lose information, in that the solution's Fourier components would not produce the saturated signal (because the np.clip operation is done on the time series during the solution process). @cmichelenstrofer and I therefore think this isn't a good idea, but I wanted to post it here for discussion and then close if appropriate.

[ryancoe]

@cmichelenstrofer - points out that capturing saturation nonlinearities would be less computationally expensive using np.clip (you would not need to have all the harmonics in your frequency array).

However, we generally feel that this is a bit of a dangerous thing, because it makes it easy to make mistakes with the post-processing. Therefore closing for now.