should be checked before almost in order to link it correctly.

I fixed some probability issue in Stopwatch.
Before this fix, the macro used to check the presence of
clock_gettime was not properly recognized. Additionally,
gettimeofday was considered as compulsory, causing problems
on bgxlc (as pointed out by Teodoro Laino).

After this fix, both gettimeofday and clock_gettime
are searched and considered as optional. If present,
clock_gettime is used. Otherwise gettimeofday is used.
If none is present, time is just not measured.

(backported from master).

@gtribello you might want to check if this is ok.
Notice that without this fix histograms accumulated in a replica exchange
simulation would include also data from other replicas, which is a bug.
I thus think this commit should be applied also to v2.0 and v2.1.

(cherry picked from commit 69578b2d)

On my workstation --no-mpi makes plumed initialization
much faster.

Manual building calls "plumed" many times. Using the --no-mpi flag
it is possible to skip initialization of MPI. On my workstation,
this makes manual build way faster.

Action::registerKeywords and ActionPilot::registerKeywords were
not called in WholeMolecules::registerKeywords

(cherry picked from commit 53a5b885)

Fixed #114

I had to remove the postproc.sh file added in v2.0
to post-process html files (in v2.1 this is done in go-doxygen)

Conflicts:
	user-doc/Makefile

I added a few tabs to the html manual:
Getting Started
Tutorials
Index of Actions

Fixes #70

internal molfile plugins are now contained in PLMD namespace

No "extern C" in molfile plugins

I fixed a bug occurring when using replicas with different
Hamiltonians (either lamdba dynamics or plumed -hrex branch).
The acceptance was wrong when using a CV that depends on charges
or masses (if charges or masses were changed in the Hamiltonian,
respectively).

related to #113