The reference module means that you can use different types of metric when
calculating the separation between the instantaneous position and a reference
configuration.

I have made a small change to SwitchingFunction so that you can get
the value of the cutoff from the tolerance.  Have also added a regtest
to check if link cells work correctly

Manyrestraints now allows you to add a bias on all the collective variables
calculated by a multicolvar.  You can thus use it to force the system have all
the atoms have a coordination number higher than a certain value and so on.

There are now a number of places where data in vessels is passed
between actions.  For instance this is done in multicolvar functions
such as local_q4 and when calculating the average value of the CV in
a particular volume.  I had done this in an ad-hoc way so there were
duplicated bits of code in different places.  I have now changed the
code so that there is less duplication of code.  There is now a single
base class which (most) of these vessel using actions inherit from.
MultiColvar functions still copy data across manually

This functionality is not complete.  In multicolvar this is
done correctly.  However, for normal colvars and the like
derivatives are still calculated even when they are not required

ignore Proline CB chemical shifts

updated manual

It is now possible to specify GRID_SPACING instead of GRID_BIN.
If both are specified, the most conservative choice is used for each CV.
If none are specified, 1/5 of hills width is used (this is
very conservative but will avoid mistakes).

I have added the possibility to use @chi1-

fix to components documentation

Linux is case sensitive

Conflicts:
	src/colvar/ContactMap.cpp

description of the components

Fix for #91, there were problems when using the keyword SUM
(components were still defined and derivatives were not allocated)
I have also added a keyword CMDIST to calculate the square
distance with respect to a reference map.
This close #90 #91.

Two bugs where there:
1. It was only working with double precision MD codes
2. It was returning energy in PLUMED units and not in MD ones

Both are fixed now

I also re-imported lapack and blas

The way it was fixed was still needing a modification in the lapack
directory which is not compatible with re-import.
Now it is fixed

For the time being dgesvd is only available with external lapack
in due course this should be made to work with internal as well.

Added routines to calculate the Moore-Penrose pseudoinverse of a matrix,
to gather a matrix from the nodes using the MPI communicator and to
do multiplication of a matrix by a scalar