Default solvent padding is often nsufficient for solute-only dodecahedron boxes.

[IAlibay]

Our default 1.2 nm solvation usually works fine when working on cubic boxes, but can lead to the following error when using dodecahedron boxes:

openmm.OpenMMException: The periodic box size has decreased to less than twice the nonbonded cutoff.

The current solution to this problem is to use a larger cutoff for solute-only calculations with dodecahedrons - usually 1.5 nm works.

This may be related to #827, specifically @mattwthompson's comment regarding the bounding box issues with small molecules.

[IAlibay]

Potential solutions discussed at yesterday's protocol-devs meeting:

1. We could have a validation step that attempts to estimate if the box will be too small. I believe (please correct me if I'm wrong!) @hannahbaumann suggested maybe generating multiple conformers and attempting to resolve the box vectors in the same way openmm does.
2. We could have adaptive settings that dynamically change the cutoff as necessary during the Protocol - which would require us transparently exposing the new settings in some way.
3. We could have context-specific defaults that change based on the ChemicalSystem composition.