I made a few more changes to setup.py in order to work with MacPorts.
This allowed me to add a temporary Portfile for py-plumed
that is now used in travis tests.

The plan is to submit this Portfile to MacPorts after we release v2.5

Notice that one test is duplicated (python/test) in order to be available
when using nosetests from python.

disabled arrayfire from travis (but I kept the code if at some point we can make it numerically more stable)

this might decrease measured coverage, but will make the job
faster and possibly capable to finish on travis-ci

Notice newer cppcheck versions are slower. In order to make the check fit
on travis-ci I had to make a few extra changes:
- better optimization flags when building cppcheck
- force a number of ifdefs in the check

Now master takes approx 25 minutes, so we have room for adding new code.

This reverts commit d190c105.

This reverts commit 14eb6054.

This reverts commit 8e6a72aa.

More efficient on travis-ci

Fixes #332

Coverage scan is now stored in a separate repository. This allows
generating the manual and the coverage scan in two different travis jobs.

Since this introduces a new travis job (for coverage scan), I removed another
job that was redundant (build with mpi and -O3; notice that we always use -O3
since v2.4). System lapack are now tested on the first job.

Fixes #348

cc @hanatok

Builds are faster on travis. Notice that all developers
recompile the code without this option, so in order to check if the
dependency options are fine we really do not need travis-ci.

macports ccache was not saved correctly

I noticed that cache is not saved if brew is removed
(apparently some travis tool requires a brew library)

Exploits 07552b60

osx8.2 is deprecated.

Problem was likely due to an upgrade of mpi in homebrew on travis

Merging debug flags and doc build (949397c0)
was a bad idea since the resulting job takes too long (doesn't
fit in 50 minutes for master branch)

I split them again, resulting in 6 jobs.

@gtribello please double check this.

There are problems with conda on travis since it expects a compiler
newer than the installed one. Namely, there were recent
failures related to unsupported gcc flags
(see https://travis-ci.org/plumed/plumed2/jobs/280964029)

goes with a5daa52d