Negative input current U not recorded correctly with snip probes

[hunse]

Just ran into this briefly, and haven't examined it in detail or tried to reproduce it in a minimal example.

When I was probing the input current for some neurons nengo.Probe(ens.neurons, 'input') with precompute=False, rather than getting negative numbers where it looked like I should, I got very large positive numbers (like 4294951040). I think these are negative numbers that are somehow being interpreted as uints instead of ints or something like that. Things work as expected if use use NxSDK probes (with precompute=True), so that can be a point of comparison.

[hunse]

I have no reason to believe this doesn't occur with V as well, but by default we don't have negative voltages, so we don't notice it.

[hunse]

Fixed in a0bfc03 and tested in df77cf8 (#169).