`Voter` trait

[avibryant]

To encapsulate the various strategies for combining predictions from multiple trees. Probably something like:

trait Voter[T] {
   type V
   def create(target: T): V
   def monoid: Monoid[V]
   def result(votes: V): T
}

with subclasses like class Plurality[L] extends Voter[Map[L, Long]], class SoftVote[L] extends Voter[Map[L, Long]], and so on.

[avibryant]

This would then leave the Error trait to take only a single predicted: T value instead of a sequence of them.