Compute collective variables in MD engine

[GiovanniBussi]

This is something we discussed long ago at the first developer meeting, I add it here even if I think the priority is presently low.

We could add the possibility to compute CVs directly in MD engine. They could be passed as: plumed_cmd("setCV pippo",&pippo); // this will be passed to plumed plumed_cmd("setCVforce pippo",&pippoforce); // this will be retrieve from plumed plumed_cmd("calc"); The MD code would be then responsible to propagate the force to the atoms. This would open to these possibilities:

• Calculation of variables depending on electronic density in DFT codes.
• Calculation of very expensive variables inside MD engine taking advantage of e.g. MD neighbor list or optimized routines.
• Implementation of a dummy MD code to test free-energy methods with simple low dimensional models (e.g. Langevin models).

[yuanchaohu]

That would be good to generalize Plumed well over various research fields