Bugfix: Save CG-matrix and reuse AMG in CG-subspace in AMG-DG backends

This patch enables to skip the triple matrix product to compute the
CG-matrix and to reuse the AMG for the CG-subspace on demand
for both the sequential and the overlapping backend.

It *fixes a bug* in the overlapping case when used together with
StationaryLinearProblemSolver with the matrix reuse turned on.

The default is the old behaviour.

See merge request !70

Cleanup: Matrix backends in boilerplate code

Replace deprecated ISTLMatrixBackend by BCRSMatrixBackend.

*We should add to the release notes and/or to the documentation how the
number of nonzeros can be cross-checked.*

See merge request !72

Skip the triple matrix product and reuse the CG-matrix together with the
AMG in the CG-subspace between calls to apply().
This *fixes a bug* and the reuse of the CG-subspace objects *is a must*
in combination with the matrix reuse in class
'StationaryLinearProblemSolver'.
The default is the old behaviour.

Skip the triple matrix product and reuse the CG-matrix together with the
AMG in the CG-subspace between calls to apply().
This additional feature can come in handy especially in combination with
the matrix reuse in class 'StationaryLinearProblemSolver'.
The default is the old behaviour.

Feature: Change parameters from outside in AMG-DG backends

As in the AMG backends for the CG-spaces, change the parameters for the
CG-subspace in the AMG-DG backends from the outside.

Adapt the test ```test-dg-amg.cc``` to use this feature.

See merge request !55

clang warns:
      moving a local object in a return statement prevents copy elision
      [-Wpessimizing-move]
          return std::move(stats);
                 ^
      remove std::move call here
          return std::move(stats);

Some random bug fixes

This merge request is just a bunch of random fixes I did while working
on the nonoverlapping mode cleanup.

See merge request !67

mpiexec sometimes fails if it is not passed the full path to the
binary. This patch fixes the problem by prepending
${CMAKE_CURRENT_BINARY_DIR}.

This code should really be moved into some kind of mixin or
something (the code duplication is horrible, but I honestly don't have
the time right now).

Also: What is this EdgeS0LocalFiniteElementMap, and does it even work?

[TypeTree] Port to new child access API

The extract_child_type<> meta function and the extract_child() function
have been deprecated in TypeTree, so this patch ports to their
replacements ChildForTreePath<> and child().

See merge request !66

The extract_child_type<> meta function and the extract_child() function
have been deprecated in TypeTree, so this patch ports to their
replacements ChildForTreePath<> and child().

[bugfix] Replace Geometry::dimension by ::mydimension

Replace Geometry::dimensionworld by ::coorddimension
They were deprecated in Dune 2.4 and are gone in Dune master.

See merge request !65

Replace Geometry::dimensionworld by ::coorddimension
They were deprecated in Dune 2.4 and are gone in Dune master.

[cleanup] reimplement PointwiseGridFunctionAdapter using variadic templates

the PointwiseGridFunctionAdapter still used the EmptyNode from typetree, which I'd like to get rid of, as it is an artifact of the non-variadic implementation of a composite node.

See merge request !64

Update convectiondiffusionccfv.hh by weighted mass matrix from transportccfv.hh.…

… Use param.d(...) for the capacity term. Added new test-transport-ccfv.cc to test this local operator.

Please see issue #34.

See merge request !59

[cmake] Issue a warning if somebody adds a test via add_test

This is just a small adjustment that issues a warning, if somebody does not use pdelab_add_test. Implemented as a proof of concept after discussion on whether it is ever useful to overload builtin cmake commands. Take it or not.

See merge request !60

Feature: Cleanup in function utilities

This cleans up the file ```common/functionutilities.hh``` with the following changes:

- rewrite function ```integrateGridFunction``` in terms of range-based for loop
- replace entity pointers by entities in class ```GridFunctionProbe```

See merge request !54

Replace Dune::array by std::array (and some other minor fixes)

Thank you Oliver!

See merge request !62

Dune::array has been scheduled for removal, as it is nowadays just
another name for std::array.

The latter one has been removed in dune-grid git master.

Bugfix: include all VTK-writers in "boilerplate/pdelab.hh"

"vtk.hh" in dune-grid includes all VTK-writers.

See merge request !52

Feature: Test for heat equation with periodic boundary conditions

This branch is Linus' work imported from
https://github.com/LSenf/dune-pdelab.git, with a minor
post-cleanup.

It provides a test for the instationary heat equation in parallel
with periodic boundary conditions in all directions.

This primarily depends on the merge-request !52. 

See merge request !56

Feature: Test for Poisson equation with periodic boundary
conditions and test for adaptivity

This branch is Linus' work imported from
https://github.com/LSenf/dune-pdelab.git, with a minor
post-cleanup.

It provides various tests for the Poisson equation with periodic
boundary conditions:

- 2D case with example taken from http://fenicsproject.org/documentation/dolfin/dev/python/demo/documented/periodic/python/documentation.html
  - using a Q1-CG discretization
  - using a Q2-CG discretization
  - using a Q1-DG discretization
  - using a Q2-DG discretization
- the example from http://fenicsproject.org/documentation/dolfin/dev/python/demo/documented/periodic/python/documentation.html trivially extended to 3D with periodic boundary conditions in x- and y-direction
  - using a Q1-CG discretization
  - using a Q2-CG discretization
  - using a Q1-DG discretization
  - using a Q2-DG discretization

The tests concerning adaptivity have been revised.

See merge request !57

Feature: New line in output of StationaryLinearProblemSolver

Insert a new line in output before solving the linear system.

This improves consistency wrt. the other solvers.

See merge request !61

Bugfix: Rename typename in darcy_FEM.hh

The name changed from Jacobian to JacobianInverseTransposed.

See merge request !53