Perf issue with multi-triggers inference on some formulas

[iguerNL]

Should re-implement the way multi-triggers are computed in Triggers.ml. It currently uses power-set, which does not scale in some situations