WIP: MultiCompartmentSystem

[wsphillips]

Fixes #3

This will allow arbitrarily large assemblies of compartments. The final benchmark will (ideally) be reproduction of models as small as Pinsky and Rinzel 1994 up to the classic Traub 1991 pyramidal neuron model with 19 compartments.

The pattern of MultiCompartmentSystem will be similar to NeuronalNetworkSystem where edges are described by Junction (similar to Synapse in networks).

[wsphillips]

The demo for MultiCompartmentSystem is found at demo/pinskyrinzel.jl. Its not an absolutely perfect reproduction, but I can achieve "reasonably close" dynamics as panels A, C, and D of Figure 2 in Pinsky 1994.

Two compartment neuron with somatic current injection (compare to Figure 2A):

For the rest of the model, I will need to implement event/time-based synapses (versus continuously integrated synapses used in the STG models)

[wsphillips]

Good enough--full model needs event-based synapses, which can live in a separate PR.