This function creates and returns a node evaluator that implements a best first search for a multithreaded search.

Nodes are evaluated according to the variable v. As long as the minimum of v is no greater than the minimum of the evaluation of each open node + s, the search continues as a depth first search. If the opposite is true, the solver postpones the evaluation of the current node and jumps to the best open node.

If a variable v appears in more than one agent (that is, more than one worker), then every copy of the variable v must have the same name.