Add internal delay capability to models

[alchemyst]

At the moment our tf object can only handle rational functions with the option one time delay. This causes an issue when calculating transfer functions like 1 + G if G contains a time delay. This issue is for allowing some kind of resolution for this so that we can calculate closed loop transfer functions for systems with delays.

[GregLaycock]

Dibs

[AlgorithmicAmoeba]

I do not mind trying to get this started. My plan would be to implement something similar to that of what matlab does. See https://www.mathworks.com/help/control/examples/specifying-time-delays.html#d120e709 for more details.

My first step will be to get this working for SISO systems.

Having looked at the current siso TF code, the potential problems I foresee are:

• calculating the inverse in cases where it is not physically realizable
• the need to write an ODE solver that handles time delay for step responses
• handling of pole and zero calculations (my current idea is to have to alternate calculations: one that uses a root finding algorithm which would be unstable, and one that uses a zero order pade like matlab: https://www.mathworks.com/help/control/ref/pole.html#d120e101850 Note: a higher order pade could likely also be used with little increased difficulty)

[ilayn]

If you keep the delay information out of the rational parts you can do operations a bit more conveniently. For SISO models output delays and input delays are the same hence might be a good place the start.