I just checked in some changes to IMP::Model. I don't think anyone was using the features (and they are relatively new).

- removed the function to evaluate the score on a subset of the particles. If you want to do that, I have a much more efficient way. Ask me.

- move the dependency graph computations out of the model since the model is already doing lots of things and they don't need to be there. get_dependency_graph is now in the IMP namespace.