Investigate the problems with running the Multi-Area-Model

[rowleya]

Determine why the multi-area-model doesn't even run in the baseline version. I suspect this is down to how SpiNNaker software currently routes populations... even if the connectivity between neurons is low, all spikes will go from source to target.

[rowleya]

First try of using a spike filter fails even on the microcircuit (see sPyNNaker#1633).

Next thought would be to try another way of getting the routers to do what we want i.e. filter more packets. It might need a combination of these techniques to work fully.