Fix/test fail dump output

This MR fixes the issue of the tester not producing output when a test executable segfaults or other system error occurs.

It also adds an option to generate a hostfile which explicitly specifies the maximum number of required slots when running MPI executables, since newer versions of OpenMPI now stop jobs requesting more than the number of available CPU cores by default (which causes our buildbot to fail).

See merge request !745

Add fix for PETSc detection on Debian 9/Ubuntu 16.04

This MR fixes a problem with detecting PETSc with later Debian/PETSc packages, which is currently preventing the Debian 9 buildbot from compiling properly.

See merge request !735

CMake updates

This MR includes a couple of updates and fixes to the CMake build system. Includes
- Remove QT4 support since we do not use it and it is now outdated.
- Change test used to identify Loki to avoid detecting a system-level subset of loki distributed by XDMF pacakge.

See merge request !722

feature/varopti_updates

New node-colouring and streamlined GetFunctional()

See merge request !738

Merge branch 'feature/jan_masterupdates' of gitlab.nektar.info:meshing/nektar into feature/jan_masterupdates
some previous merge

Feature/bessel

This MR introduces Womersley boundary conditions for pulsatile flow in straight pipes. To provide the appropriate boundary conditions a file specified in the 
USERDEFINED BC string is appended with a colon and then a filename, i.e.
            <REGION REF="0">
                <D VAR="u" USERDEFINEDTYPE="Womersley:WomParams.xml" VALUE="0" />
                <D VAR="v" USERDEFINEDTYPE="Womersley:WomParams.xml" VALUE="0" />
                <D VAR="w" USERDEFINEDTYPE="Womersley:WomParams.xml" VALUE="0" />
                <N VAR="p" USERDEFINEDTYPE="H" VALUE="0" />
            </REGION>
where the filename WomParams.xml contains the setup details. Ideally we would like  a better solution than using the colon extension and an issue has been raised on this.

See merge request !472

# Conflicts:
#   CHANGELOG.md

Fix FillWhiteNoise

This MR fixes a bug in `Vmath::FillWhiteNoise` which causes it to alternate between only 2 values.

Instead of trying to pass a unique seed on each call of FillWhiteNoise, I decided to partly recover an old implementation where `seed` was an static variable ( therefore making consecutive calls part of a single larger random sequence). From what I understood, this had been changed to allow specifying different seeds for each MPI process. With this in mind, we can still pass an optional seed to replace the stored value.

An example of how to use this new logic is to use (the negative of) the rank as seed in the first call to FillWhiteNoise to decouple the parallel processes, and not passing any seed in the subsequent calls . Alternatively, we can still pass unique seeds on each call.

See merge request !718

# Conflicts:
#   CHANGELOG.md

feature/mesh-spacing

The applies changes ask for my spencer after the madrid tutorial, BL resolution and variable names etc.
It also fixes the inverted boundary layer problem.
Finally, allows the variational module to be compiled without meshgen.

See merge request !736