[BUG] Periodic boundary conditions for OpenMM restraints

[lohedges]

At present, the restraint forces that are created during the Sire-to-OpenMM conversion are set to use periodic boundary conditions. For bonded restraints, this doesn't make sense since a restraint could be satisfied in different ways depending on whether it is measured across the boundary or not, e.g. a molecule could extend, or contract, e.g. in a ring-breaking simulation.

I'm not sure whether we want to update all of the restraint classes in Sire::MM to add a use_pbc option in the constructor, or simply disable periodicity in the conversion layer. The former is more general, but requires a lot of updates and changes to the streaming operator versioning for all restraint classes. One complexity is the "distance" restraint, which is also used for "bond" restraints. I imagine the desired periodicity is different depending on whether you want to restraint a bond, or keep an atom close to some reference position.

[lohedges]

It might be easiest to restrospectively call the setUsesPeriodicBoundaryConditions() method on the OpenMM force in the Sire Python layer.

[lohedges]

I'll just monkey-patch the Python wrapped restraint objects, i.e. add a .use_pbc attribute, then check for this in the conversion layer and pass through via the map.