Infrastructure for portfolio solving (#8709)
We want to replace the set of competition scripts by an internal portfolio mechanism. It seems that this portfolio mechanism is best implemented from scratch in the driver instead of using, e.g., --tlimit or --tlimit-per.
This PR introduces the infrastructure for that, implementing the following approach:
first parse the input until a logic has been specified; if more than one configuration is to be used, parse the remaining commands into a vector; for every configuration, we fork the main process, and in the child process configure the solver and replay the vectors.