This is a huge commit, applying indentation to the whole
code. Notice that this commit is made on master branch
after having merged with -s ours from branch v2.3

See #215

This is a huge commit, applying indentation to the whole
code.

See #215

I added a script located in src that applies astyle to
all source files and add them to git. This script can
be invoked as `make astyle`.

File .astyle.options contains the astyle command line options

See #215

Prior to this change a misuse of the LOWMEM keyword could cause segfaults.
Now these cases crash with an error.

MOVINGRESTRAINT with TORSION

It is difficult to remove the main timer, but at least I
added a timer that makes sense

Fixes #194

[makedoc]

[makedoc]

- Vim syntax for the awk file
- Some information is given once per file now.
- Registration of vessels and metrics is also logged.
- Check that .h files do not include:
  * inc files (they are temporary)
  *  external .h file (they might be not available when using plumed library).
The latter is only notified now, since it must be accepted for mpi.h
- Separately store documented actions/cltools/other.
- Check that module type is correct.
- Extensive doc for each check added.

As requested by @gtribello (see #221)

Also allow a hybrid syntax such as
AT={1 @multi:{5 6 7} 2}

In this case the first and third arguments are the same across replicas,
whereas the second is 5/6/7 on the three replicas respectively

Implemented syntax to allow simpler syntax with multiple replicas
using an almost identical input file.

Closes #221

This is done in the following way:
- Special value of biasfactor for non-well-tempered is -1.0
- When biasfactor=1, -F is stored instead of V

In this manner the only change visible from outside is
that HILLS file produced with non-well-tempered metad
have biasfactor=-1

Addresses #220

Allows specifying multiple BIASFACTORs in a single file.

In addition, it triggers a citation to "altruistic" paper
by Spiwok et al when using METAD with RECT and WALKERS_MPI.
Notice that in this case the free-energy estimator is shared
rather than the bias itself, so as to allow "altruistic" updates
when walkers have different biasfactors (as in RECT).