I also fixed made the data collecting class is aware if we are calculating
a distance.  If it knows that we are trying to calculate a distnace
it returns something with which it can calculate a distance and not
the position stored by COLLECT_FRAMES.  I am not sure this is necessary
and if it will be ever used but I guess it is better to be safe than
sorry

You can now use COLLECT_FRAMES to collect pdb data from the trajectory
and output a pdb file that contains the underlying atomic coordinates
in pdb format even if you are not using the rmsd distance to measure
the dissimilarity matrix for dimensionality reduction.  Also fixed
a number of bugs.

These classes are no longer necessary or useful in fact
they were just confusing.

These classes are no longer necessary or useful in fact
they were just confusing.

The data is now passed via a pdb object.  This means we can now use
the same infractructure for setting up the ReferenceConfiguration as
we use in setting up the ReferenceConfigurations in paths and so on.
Projections meanwhile are attached to the ReferenceConfigurations
when they are required for printing.  So now a reference configuration
can be the pair of objects - i.e. the projections and the high dimensional
configuration.

I have moved mds from the analysis module to a separate dimred module
as I think it is best to keep this with the other dimensionality
reduction algorithms.  It is now default off though

The old analysis class was reaching a stage where it was not fit
for purpose.  This is thus a total redesign in order to allow us
to write better algorithms for dimensionality reduction that are
flexible and easy for users to use.  I have removed some functionality
with this commit but I will re add it in subsequent commits
Also added information on analysis classes to developer documentation

[makedoc]

[makedoc]

[makedoc]

[makedoc]

I did it only when using the advanced syntax (SWITCH={...}).
I changed the input in some regtests to comply with
this change.

(cherry picked from commit b6ea36c1)

Actually, this was a workaround I found on the web for MSC,
but looking at the code I don't think it should work.
Probably, it has never been tested, so better remove it.

Now all the steps are guided

@gtribello please check

Chnages were made in a number of places to (a) ensure that this
functionality was available in a large number of multicolvars and
(b) to prevent duplication of functionality between the new vessels
for the maximum and minimum and the old colvars for calculating functions
of all the multicolvars

I removed the NOPBC key from manual (it is not used)

[makedoc]