Make extendedLagrangian and ABF work with vector CVs

[jhenin]

This is a planned enhancement.

There is no fundamental reason not to have this. We need a general approach to address a collection of colvars as a vector of their scalar elements.