Make it easier to access model evaluation results

[PGijsbers]

Currently you can access them from evaluations.log or with gamaobject._evaluation_library.n_best(n, with_pipelines). The first is a little cumbersome if you still have access to the gama object, the second is unintuitive when you use BestFitProcessing as you then have to set with_pipelines=False explicitly. It would be great to have:

• something like a gamaobject.leaderboard property which returns a dataframe with results of each evaluation
• modify n_best behavior to return the best evaluation objects that are relevant to the initial configuration (if no pipelines are ever meant to be stored, the with_piplines=True default makes no sense)