Before: each layer(neuron) had its own spike buffer
  Now: A single buffer stores all spikes (post, weight, arrival_time) as a list of triplets

In a scaled down version, this resulted in simulation time decrease to about half:
  Before: about 400s secs for 1 sec simulation
  Now:    about 160 secs for 1 sec simulation

  Before: array (would consider only the first synapse arriving to a specific neuron)
  after: use for loop to account to multiple synapses arriving to the same neuron