Idea for Successive Halving to avoid repeated re-evaluations of the same pipeline.

[perib]

Currently, successive halving reduces computation by subsampling the number of rows. Every time the budget increases, the existing parent population is re-evaluated at the next budget, which may be inefficient. Another approach could be to make budgets related to CV folds. We could use only one fold to evaluate the first N generations. Then evaluate two folds when the budget increases. In this case, the parent population wouldn't have to be re-evaluated from scratch; it would just be evaluated on the next fold of CV.