Bias-exchange metadynamics implementation

[giacomofiorin]

This PR contains an implementation of bias-exchange metadynamics (https://doi.org/10.1021/jp067873l) based on the NAMD Tcl scripting interface.

Although most of the code is Tcl, the following significant changes are introduced to the C++ source code:

• The value of the "active" flag of the biases is being written to the state file; this is needed because all biases are defined for all replicas, but only one is active at a time. Therefore, we need to carry over the active flag when checkpointing.
• The value of the "active" flag can be reported on the trajectory.
• The field name is now a member of colvardeps, and could later be moved to the AST (see [RFC] More band-aid fixes to Colvars #700 (comment) and related comments).
• Allow updating and reporting in the scripting calls the energy and forces of a bias, but without actually sending those (used to decide the exchanges).
• Adds a neutral bias, which simplifies greatly the inclusion of replicas with equilibrium sampling in the BE implementation.