pip 10 moved several interfaces to an internal package, causing our pyversion.py script to fail.
This patches fixes the issue by getting installed distributions directly from setuptools, as
suggested in https://github.com/pypa/pip/issues/5243.

See merge request !538

When the virtualenv support was added to the build system,
https://bugs.launchpad.net/debian/+source/python3.4/+bug/1290847# made us bail out
completely on building a virtualenv with preinstalled pip. The situation has improved
a lot since then; even Ubuntu 14.04 gets this right now as long as you install the
right packages. So let's be nice to our users and try whether installing pip works
before forcing them to do it manually or use `get-pip`. If a user insists on `get-pip`
by setting `DUNE_PYTHON_ALLOW_GET_PIP`, we skip this additional step and keep using
`get-pip`.

This shouldn't break anything, so I'll backport it to 2.6 as well, as it should make
it much easier to use the virtualenv in CI images (e.g. for
`dune-pdelab-systemtesting`, which requires `dune-testtools`).

See merge request !536

Way back when we first integrated a virtualenv into the Dune build process, there was a nasty bug in
several Linux distributions that made it impossible to get a standard virtualenv with pip installed
inside of it, so we decided to completely skip the default installation of pip into the virtualenv
and instead download get-pip to install it. This is not always appropriate, so now the flag
DUNE_PYTHON_ALLOW_GET_PIP has to be enabled for it, but without it enabled, we still force users to
manually install pip into the virtualenv.

Meanwhile, pretty much all distributions managed to get their act together (even Ubuntu 14.04 now
correctly installs pip into the virtualenv if the required packages are installed). In order to make
life easier for our users, this patch changes the default behavior: It now tries to create a
virtualenv with pip installed first, but if that fails, it reverts to the old behavior.

dune-ctest: do not print messages from failing tests twice

Closes #126

See merge request !535

CTest already prints the messages once.

Closes: #126

GCC 6.3 errors out with an ICE or simply hangs when instantiating a certain variant of `operator^=` on
`BitSetVectorReference`. This patch works around the problem by forcing the actual XOR instruction
into a separate function for that compiler. The workaround is probably horribly slow, but it
shouldn't have any impact on unaffected compilers.

This probably needs backporting.

See merge request !528

GCC 6.3 errors out with an ICE or simply hangs when instantiating a certain variant of operator^= on
BitSetVectorReference. This patch works around the problem by forcing the actual XOR instruction
into a separate function for that compiler. The workaround is probably horribly slow, but it
shouldn't have any impact on unaffected compilers.

This probably needs backporting.

This updates the CI configuration to use the new Docker images and their different toolchain
configuration. Note that this causes a build failure on Debian 9 with GCC 6 due to an ICE or
a compiler timeout. I'll push a separate MR to fix this issue, as the fix will need to be
backported.

See merge request !527

[CI] Optional expensive tests.

See merge request !523

Spelling in comments.

See merge request !525

Depending on the Simd library, this leads to arrors about list-initialization.

Depending on use, use either the previously introduced `Simd::broadcast<V>()`
function, or in the testsuite to the incantation directly to avoid test
interdependencies.

[Simd] Introduce broadcast<V>(s) function.

See merge request !524

This adds build that runs expensive tests, including Vc on with at least AVX.
The build is allowed to fail to compensate for the possibility of no runner
beeing available.

Feature/float128

See merge request !518

[bugfix] add missing include for `std::integer_sequence`

See merge request !521

[bugfix] VariableSizeCommunicator fixes

See merge request !520

Otherwise we are trying to receive more requests than are there and get
dubious exception because of resizing vectors with MPI_UNDEFINED.

MPI_Wait* and MPI_Test* will ignore these.

Indeed we cannot assume that we always have to send data along one.

[SIMD/test] Do not access indeterminate simd vectors with lane().

See merge request !519

Depending on implementation, this may require bitmask operations on the
indeterminate value, which requires an lvalue-to-rvalue conversion, which
results in undefined behaviour except in some very special cases.

Add helper function unpackIntegerSequence

See merge request !495

[SIMD] Test the index vectors the same way we test other vectors.

Closes #125

See merge request !516

In the tests for operations we check whether an operation is supported on
scalars, and if so, we also test is in simd types.  This suppresses tests for
e.g. `%` on simd types made up of floating point types.  Before this patch we
sometimes did the check on proxies instead of scalars.  This could lead to
tests be incorrectly not suppressed.