core: Consider rotational kinetic energy of virtual particles.

[fweik]

Fixes #3596.

This is a fix closest to the existing behavior. I think there should be a discussion here how
to handle active/inactive degrees of freedom, and if this should be consistent between
translation and rotation, and how it should interact with the virtual sites relative feature.

Description of changes:

• Don't drop rotational kinetic energy of virtual particles.

[RudolfWeeber]

I think there should be a discussion here how to handle active/inactive degrees of freedom, and if this should be consistent between ranslation and rotation, and how it should interact with the virtual sites relative feature.

Yes. The current behaviour with regards to VS relative and rotation is hard to understand. My current understanding is that, the propagation for rotational and translational degrees should be controllable independently and per particle. E.g., users could specify * Posiotn propagation: Relative to other particle * Rotational propagation: Rotational Langevin eqn From that information, it should be derivable whether the dof is ‘active’ or not. This would be part of #4015.

[fweik]

In my mind, in an ideal world you'd be able to add or remove degrees of freedom to a particle/point, and they would themselves know how they are propagated and what their energy/momentum contribution is. But probably such a fundamental design change will be hard to pull of, so I think what you are proposing is a reasonable direction. We should discuss beforehand who (in terms of components) should be responsible for that, and which choices yield that best separation of concerns, given the circumstances.

[kodiakhq]

This PR currently has a merge conflict. Please resolve this and then re-add the automerge label.