Controlling the Wallclock time by limiting the Number of individual evaluations

[HenrikMettler]

Currently the evolve function is limited by the number of generations or attaining a fitness goal (min_fitness). However the number of generations is only loosely related to the number of fitness evaluations, since only individuals with a novel phenotype (new configuration of active genes) are evaluated. In a qualitative comparison study for cgp's evolutionary mechanisms (https://ieeexplore.ieee.org/document/6815728/?arnumber=6815728) Goldman et al. benchmark their parameter tuning on the number of fitness evaluations. I would argue that this could be a reasonable approach for us as well, since presumably in most of the implementations:
T(evaluating_individual) = big
T(calculating_genome_output) = small
Therefore allowing the user to define a maximal number of fitness evaluations might simplify the control of a maximal simulation duration.

[mschmidt87]

Makes sense to me. This could be an additional keyword argument for the evolve function which can be set instead of max_generations.

[jakobj]

an additional keyword makes sense, but i wouldn't make the two exclusive: just break whenever one of the two reaches it's maximal value.