Adapting existing deme expander and optimizer implementations

[K1rubel]

Detailed Description

Major Tasks

1. To integrate strategy scoring implementation into the transform function which is responsible for converting an instance to tree and scoring the tree against an adversary tic-tac-toe strategy. This is a procedure called by the main optimization algorithm, hillClimbing.
2. Using caching mechanism for behavioural scoring used in demeToTrees as well as transofrm in the hillClimbing procedure. Refer to this issue for more description.

Other minor tasks would be

1. Cleaning up the existing hillClimbing in a way that strips it off parameters that are not relevant for this specific case.
2. The same thing with deme expander