SP replacec bumpUpWeakColumns with dendritic death/creation

[breznak]

Once #537 is merged, test synaptic and dendrite death in SP.
The method SP::bumpUpWeakColumns() could be replaced with death of the unused segment.
And generete new one?

• when? Each time a seg/col is removed, randomly, ....?
• where to generate a new col?

Related #288

[breznak]

[on removing empty dendrites]

Also I think that this code branch will be reached very rarely, because if there are few synapses on a segment, then it is unlikely to activate, and adaptSegment is typically only called on active segments.

#537 (comment)

That is true, considering implementing "random" parallel energy-monitor (aka cell killer).

• cells do not have to do this during normal computation
• killer will run in its own thread: when finds too low synapse/segment -> remove it.
  • how to make this work with deterministic algorithms? Deterministic Algorithms #194
  • even low-energy/low activation segs/syns get removed, as they do not need to activate anymore, but the death visits them.