Test negative time

[JamesPHoughton]

I've come across a few models that start time at -1, or some other negative value, so that at time 0, they introduce a disturbance, and track the response from the time of the disturbance. Its nice, and clean. A good example is Lyneis and Sterman 2016

But, we should check that it works with all of our special functions that also depend on time: step inputs, ramps, pulses, etc.