Added max population setting so that the population expands to max size at pop 2

Added gaussian log likelihood functions
Now we generate a new model at the beginning of each population
Fixed some bugs in the feature generation, projections now must have two features and it does not get stuck on collinearity testing