Investigate/remove bound checks

[EmirDe]

Maarten previously discovered that removing bound checks increased the solver performance significantly (+50% speed?), based on some instances. Most likely this is all tied to some small part of the code, e.g., nogood propagator and/or assignments. That is a sizable improvement for hopefully not so much effort, so it makes sense to think about it. Some points:

• do a more thorough investigation on a broader set of instances, just to see whether this increase is consistent.
• isolate where the improvements are coming from, e.g., if we disable bound checks just in the nogood propagator and/or assignments, does that account for all the gains?
• See if we can avoid the bound checks easily (probably not).
• discuss the trade-off between unsafe and performance. We could have our own pumpkin checks at moderate level whenever we removed the bound check.