Fixed the initialization of some static data by treating them as
static vectors and initializing them with a lambda function.

In principle, this rules out possible problems arising when multiple threads
call the decode function simultaneously for the the first time.

The vectors are then declared as const for clarity.

opMap is now constructed in a single statement. This avoids
problems possibly arising with multiple threads accessing for the first time
to this method simultaneously.

opMap is then made const for clarity.

It is not completely clear what would happen with multi threads.
Since the function is not used anywhere, I just remove it.

cc: @valsson

* copying and adapting the FaMetaD code from the old version, still need to be tested

* forgot keywords

* fixed a few things in the FaMetaD

* improved parsing, error msg, and log printing

* moved updating of pace to update()

* astyle fix

* made FaMetaD work with current acc cacluations

* regtest for FaMetaD

* added printout of reading in of acc value

* fixed a newline error in acc printout

* added manual info for FaMetaD

* astyle fix

* removed not needed values from colvar file in rt

* fixed a problem with restarting of FaMetaD

* added rt for restarting of FaMetaD

* add notes to the manual on restarting and acc factor

* fix for restart regtest for fametad

* added warning if restarting and not reading in acc from file

* astyle fixes

* Added a check to make sure that combing FaMetaD and MPI mwalkers is not allowed

* changed error call to be consistent

Allows embedding a separate PLUMED instance.

Using auto when creating a unique_ptr

Using c++11 to initialize the map instead of calling a function everytime

This is to avoid potential clashes.

I removed many cases that are not in practice useful. In particular,
it won't be anymore possible to:
- Link statically PLUMED if both "ar" and "ld -ro" do not work correctly.
- Link runtime PLUMED is "ar" does not work correctly.

Both statements are true only for installed PLUMED. Not installed plumed
still uses object files.

(see #352)

All atoms were erroneously requested also after the first step.

Allow COMMITTOR to stop driver.

when using DEBUG, detailed timers were always shown by mistake.
Now they are only shown when using flag DETAILED_TIMERS

I added a configure flag --disable-absolute-install-name
to allow OSX shared libraries to behave such as Linux libraries

This will need DYLD_LIBRARY_PATH to be set, which is not a good idea.

I have to add some documentation.

On OSX, it is standard to hardcode the absolute path of a library in its install_name.
On Linux, the standard behavior so to only hardcode the soname and let
the system find the library. This actually requires setting LD_LIBRARY_PATH.

I added the possibility to use ./configure --enable-absolute-soname
to make Linux behave such as OSX.

I took the occasion to simplify considerably the linking of shared libraries.
Now the same command is used on Linux and OSX.

- do not set LD_LIBRARY_PATH on OSX
- set LIBRARY_PATH also when dynamic libraries are disabled

Allow avoiding rpath-link. In addition, it will make
absolute sonames on linux possible (later commit).

removed sentence about NOSPLINE

https://groups.google.com/d/msgid/plumed-users/4bdade66-5dde-4748-b3c2-3c6155906b69%40googlegroups.com?utm_medium=email&utm_source=footer

added syntax highlight in an example

all -> allatoms
physical -> mdatoms
virtual removed

See https://github.com/plumed/plumed2/pull/372

This will report an error instead of just crashing when an
incorrect symbols is used

@ hydrogens
@ nonhydrogens

See #195

See #195

@ water
@ ions

See #195

@ nucleic (DNA or RNA)
@ protein (protein)

See #195