Memory requirement checking in ManyCore ends up in a counter intuitive behavior
Currently, if the available memory is less than 500MB, we only use one solver in ManyCore (thus ManyCorePB).
That behavior is quite counter intuitive because there are some use cases (e.g. debugging in p2) where it is nice to use a normal solver in one "thread" and a StringSolver on the other "thread". This is even worst since we now allow to provide a list of solvers directly on the constructor.
It is better to remove that check, at to let the user decide exactly how many solvers to use.