Add autotuning functionality to simulator

[omarshafik]

The simulator should support monotonic autotuning parameters for quantities of interest. Here, I add a simple & generic autotuning functionality.

[maedoc]

This is a pretty good first take on the idea. I will think about some of the concrete use cases. For instance, we might take a model like the reduced wong wang, and increase each region's excitability until the firing rates are all matching, then continue to increase them all until the functional connectivity match is maximized.

in any case, I think the implementation is nearly sufficient; in the case where we're trying to maximize some measurement, we'll eventually take a step and the value will go back down? So it might make sense to trace the the tuning, and keep the best value.

What do you think?

[omarshafik]

makes sense. thanks for the hint!
I changed the implementation to introduce some logic to allow more flexibility in the tuning algorithm. Now, I use PID, but I think the logic can be extended to include other algorithms as well.
let me know what you think..

[omarshafik]

one thing to note here though is that delta is assumed to return a scalar value. This however, can be changed as simply as passing an n-dimensional threshold (and changing the type hints). I believe the rest of the logic should hold.. I could add that in a later commit if needed :)