Experiments on better implementing of active/passive

1) both sets are represented as multiset; lead to speed-down (aparently,
   because of the fact that it is harder to work with this data structure
2) both sets are represented as in Active. Implements better; however, 
   it seems that the fact that passive is ordered by the leading literal
   is more essential as lead to less number of resolvents
3) fair (but inefficient) subsumption algorithm
