Added a new cmd "setRestart" that allow setting restart
from MD engine.

This setting can be overridden by PLUMED using:
RESTART
or
RESTART OFF
to respectively switch on or off restart from PLUMED input.

Notice that this would break compatibility in the sense that
it would turn on restart automatically without the user
specifying it.

I increased the APIversion to 3 so that MD engines can
check if PLUMED support this keyword before using it.

This fixes #100

In fortran counting starts at 1 whereas in c counting starts at
zero. I have thus modified the passing of gatindices so that if you
pass from fortran you can subtract one from the index

When specified, the update() method is only invoked
within this time range. This allows to perform some tasks
in a single simulation, e.g.:
Equilibration + Metadynamics + umbrella sampling refinement
with analysis of the last part only.

Currently this is an optional flag for all the actions.
I am not sure this is a good idea (currently update() method
is only used in some ActionPilot), since this will make
the keyword appear in every manual page.
However, since the update method is in principle available on all
the actions I preferred to add these keywords everywhere.

Fixes #65

Now also looks for env vars:
PLUMED_NUM_THREADS
PLUMED_CACHELINE_SIZE

(cherry picked from commit 9c8bb507446521ff03b4e6581a99c6d183a2ab54)

Conflicts:
	src/core/Atoms.cpp

Work is correctly computed for METAD and MOVINGRESTRAINT, and subtracted
to the effective energy drift

(cherry picked from commit b450f1525a218e8181f3933779bb17e550384f61)

Conflicts:
	src/bias/MetaD.cpp

I added the possibility to pass kbT from gromacs to plumed.
To allow compatibility between old and new versions, I used the API
number. Namely: gromacs checks that plumed API>1; if so, pass the
temperature.

KbT is only used in METAD so far (not in HISTOGRAM)

Addresses #35

This is the parsing I showed you in belfast

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

Mode is redundant since:
* read or write can be decided depending if file is IFile or OFile
* append/non-append is decided based on RESTART option

Plumed paper has now a volume and page

Fixes #55

when a domain of domain decomposition is empty.

Check for NULL pointers is postponed to class Atoms and is done only if
there are local atoms. This fixes the case where GROMACS put zero atoms in
 a processor and assign a NULL pointer to positions

I think this is more coherent with other cmd() calls.

I set it to 2 instead of 2.0, so that it will not be necessary
to update it

cmd("useNaturalUnits") was erroneously crashing

Still we should write some doc. Also, there is not regtest for this...

I moved the reference together with the others (using cite()) so that
it is more visible. I also updated a bit the description

uninitialized variable (from last commit)

(suggested by Paolo Raiteri)

ENDPLUMED was not correctly understood

Can be activated with DEBUG DETAILED_TIMERS, and provide action by action
timing. Useful for optimization

Gromacs patch has been fixed (this is compulsory for big Endian machine)

Energy was not calculated at every step. Now, gromacs ask to plumed
whether energy is needed at that step and calculate it accordingly.

I also updated the patch to 4.6.1

When the log file is chosen by name (cmd("setLogFile"...))
there was a problem due to an apparent double initialization
of log file. Now, logfile is open when cmd("setLogFile"...)
and, in case it is still not open when cmd("init") is called,
it is opened to stdout

This I found to be helful when debugging on the cluster with buffered output.

When using IFile or OFile only, it is sufficient to include
one of the two header files.