Add an MC minimization stage in regularize for bad crds

When there are missing atoms in a PDB file, we end up with many
undetermined coordinates in the population. RMSD minimization
can not help us get good values for these. When a ProFASi simulation
starts with such a structure however, we have to fill in all atoms.
The ones originally missing in the PDB will have their XYZ coordinates
calculated based on nominal default values for those coordinates,
which may put them in overlapping positions with other atoms. This
leads to very high energies when we start our energy minimization
rounds (in billions of k_B T). When the energy is that high, any
MC update which resolves the clash, no matter how detrimental to
other energy terms and the RMSD, will be accepted. In those cases,
"regularize" fails to produce any reasonable structure. Here we
try one round of MC based energy minimization using only the bad,
or indeterminate, coordinates. If the clash can be resolved like
this, we will have a much better starting position for energy
minimization at the same RMSD value as obtained in the first stage
of regularization: the pure RMSD minimization. As it turns out,
the result is a bit mixed. In some structures it does wonders, in
others, it has no effect while consuming some more time.