Calculating LJ-interactions using Particle mesh Ewald

Calculating LJ-interactions using Particle mesh Ewald

The calls to ewald_LRcorrection and gmx_pme_do in force.c have been
changed to also handle LJ-PME. The unused implementation with
tables in ewald_LRcorrection have been removed, but would in principle
be faster. But since the Verlet kernels won't call the routine and
group kernels are on a limited lifespan we do not do this optimization.

The switch-statement over the different electrostatic interactions in
force.c are now handled with if-statements.

The grids in PME have been modified to now be larger "arrays of grids",
where the first two are reserved for electrostatics(A and B-state) and
the remaining ones are for LJ-PME.

The code in gmx_pme_do have been changed to also handle LJ-PME.
Previous loop over 2 states in electrostatic PME now also loops
over two additional states for LJ-PME with geometric combination
rules. At the end it contains a separate part which loops over
the various terms for the case with LB combination rules.

Added a new enum for the parameters to be communicated in pme_pp.c

A lot of small changes in various other files mostly to fix
variable-names and small LJ-PME related details. Kernel-generating
scripts for AVX are modified to access fr->ewaldcoeff_q instead of
fr->ewaldcoeff, which doesn't exist any more.

Turning on LJ-PME in simulations is just a matter of setting
vdw-type = PME in the .mdp-file and setting "lj-pme-comb-rule"
to either "LB" or "geometric"

Manual updated with new section called "Long-range Wan der
Waals interactions" which includes the old "Dispersion correction"
and a section for LJ-PME.

Uncrustified files and updated copyright headers

TODO:
1. Full functionality for LJ-PME with free energy calculations, the basic
   framework is there but currently it gives a gmx_fatal
2. LJ-PME with verlet-kernels, currently gives gmx_fatal
3. Integrate with PME tuning

Change-Id: I6395b0a6bcb887b30311143415fde5496c6f3d38