feature importance determines parent pressure for selection

[lacava]

add this configuration option.

[lacava]

steps

• adding self.weight_parents = True or False (default False)
• edit variation.py
• set weights = feature scores, normalized between 0 and 1
• replace copy.deepcopy section with numpy.random.choice with p = weights

[lacava]

commit b6310de adds weighting option weight_parents to accomplish this.