Idea: precompute var distributions

[pvdz]

Var distributors are currently applied during runtime. So every time a new step in the search tree is created the var distributor is applied over and over again. What if instead of searching we could precompute the list of vars in the order as specified by the distribution(s) picked?

• var distributors are a bit more complex than value distributors. there's an intricate system of fallbacks if no suitable var can be chosen under the current distributor
• you would still need to search through the ordered list for the next unsolved variable according to the current node of the search tree. Is that cheaper? Consider large sets of 10.000+ vars.