Repeat of spike-like-event conditions in synapse code flawed

[neworderofjamie]

In some cases e.g. "$(V_pre) > -50.0" this is likely to work but e.g. a condition that is only true before the spike reset code is run or one triggered using an extra global parameter will fail as the condition won't be true when it's re-checked next timestep (/after axonal delay) in the synapse code. There is also a much more minor issue that the threshold condition code is not considered when determining which neuron variables require queues in ModelSpec::finalize.

I think the best fix is for each spike-like-event condition (give or take some merging which is already done) to generate its own array of events.