Lifted unit propagation should detect when it will be too expensive

[Broes De Cat]

For very large structures, lifted unit propagation might run very long and indeed derive very little (remember, anything it finds out will eventually be found out when we ground it too). So maybe in fact we do not want to run lifted unit propagation on such instances, but on the other hand it would have if it maintained a cost of running the propagation compared to how much information it will derive.

[Broes De Cat]

Removing milestone: Later (automated comment)