Circumstances forced me to finally learn what the BCRSMatrix 'implicit'
build mode really does, and where its limitations come from.  I had
to read the code for that.  As it turns out, the documentation of the
mode is quite misleading.  My main gripe is that the code uses two
different auxiliary data structures, but the documentation calls them
both 'overflow'.  This problem even extends to the variable naming
in the code.  Behold: the parameter 'overflowsize' is NOT the size
of the member 'overflow'.  It doesn't even have anything to do with
that member...

To improve the situation, this patch does three things:
a) It rewrites the documentation.  In particular, the 'overflow area'
   is now clearly distinguished from the 'compression buffer'.
   The latter is a new word I introduce.
b) It renames the BCRSMatrix method parameter _overflowsize to
   compressionBufferSize, because that is what it is:  That parameter
   has nothing to do with the 'overflow' data member, or even the
   concept of 'overflowing' in general.
c) It renames the exception 'ImplicitModeOverflowExhausted' to
   'ImplicitModeCompressionBufferExhausted', for the same reason.
   This is the only interface changes.  The code keeps the old
   exception for backward-compatibility, but makes it trigger a
   deprecation warning.

Merge branch 'feature/cholmod-common' into 'master'

ref:core/dune-istl\> This allows other implementations that are using the
istl-cholmod wrapper to set options and other parameters of the cholmod
solver.

See merge request [!366]

  [!366]: gitlab.dune-project.org/core/dune-istl/merge_requests/366

Merge branch 'bugfix/cmake_guard_condition_overlappingschwarztest' into 'master'

ref:core/dune-istl\> This MR fixes the CMAKE_GUARD condition for the
overlappingschwarztest.

See merge request [!367]

  [!367]: gitlab.dune-project.org/core/dune-istl/merge_requests/367

This allows other implementations which are using the istl-cholmod wrapper
to set options and other paramters of the cholmod solver

Merge branch 'feature/parallel-solver-factory' into 'master'

ref:core/dune-istl\> This is my proposal for supporting the parallel solvers
in dune-istl with the solver factory.

Summary of the changes:

-   The parallel operators let one get the underlying communication object.
    This is needed to pass it to the parallel preconditioners
-   The solver factory passes the Operator instead of the matrix to the
    preconditioner factory. Thus AMG can be constructed without creating
    another operator from the matrix and we can determine whether we are
    parallel or not by the operator.
-   For the parallel solvers we need to initialize the factory with the
    corresponding parallel operator.

Note that further refactoring might be required to support e.g. the
grid-communcation-based parallelization approaches sometime used in PdeLab.

This is one part of the work proposed in [#82]

See merge request [!349]

  [#82]: gitlab.dune-project.org/NoneNone/issues/82
  [!349]: gitlab.dune-project.org/core/dune-istl/merge_requests/349

Currently the hierarchy building does not work for complex.

```
dune/istl/schwarz.hh:280:12: warning: struct 'ConstructionTraits' was previously declared as a class [-Wmismatched-tags]
    friend struct Amg::ConstructionTraits<BlockPreconditioner<X,Y,C,P> >;
            ^
```

It is the linear operator anyway. Hence we can deduce the rest.

We need to use the correct discretization for this. Using
the overlapping one we are getting nans.

Of course all ranks need to participate in the computation.

This is a preparation for the parallel solver factory and will allow
to construct AMG from the ptr passed to the factory.

Merge branch 'feature/remove-more-deprecated-stuff-after-2.7' into 'master'

See merge request [!364]

  [!364]: Nonecore/dune-istl/merge_requests/364

It fails when SuperLU or one of the SuiteSparse
libraries are missing.
We should develp a more fine-grained solution.

Only SuperLU 5 is now supported.

Merge branch 'fix-typo' into 'master'

See merge request [!362]

  [!362]: Nonecore/dune-istl/merge_requests/362

Merge branch 'bugfix/ConstructionTraits' into 'master'

ref:core/dune-istl Redefine all remainig class definitions as structs.

See merge request [!361]

  [!361]: gitlab.dune-project.org/core/dune-istl/merge_requests/361

Merge branch 'fix/amg-verbosity-default' into 'master'

ref:core/dune-istl A default verbosity of 2 clutters everything with a lot of
output, per default.

See merge request [!360]

  [!360]: gitlab.dune-project.org/core/dune-istl/merge_requests/360

Merge branch 'feature/solver-factory-naming-schemes' into 'master'

ref:core/dune-istl Apply naming schemes to solver factory function, improve
some internal variable names and correct some spelling and grammar mistakes.

The most severe issue fixed:

-   the functions default_preconditoner_BL_creator and
    default_preconditoner_creator didn't comply with the naming scheme and
    also had a spelling mistake: missing i in preconditoner.
-   in the usage of registry_put the name didn't make it clear that it is a
    macro. Also the macro should definitely be prefixed with DUNE to avoid
    conflicts with other libraries -\> changed to DUNE_REGISTRY_PUT.

@gruenich Should probably be fixed before the release 2.7 so we don't need to
deprecate the names.

See merge request [!350]

  [!350]: gitlab.dune-project.org/core/dune-istl/merge_requests/350

Merge branch 'feature/factory-determine-amg-criterion' into 'master'

ref:core/dune-istl Note that this changes the default criterion to the
symmetric one that uses FirstDiagonal for computing strength of connection.
Previously it was the unsymmetric one with rowSum.

Note also that the new default will not work for complex.

Closes [#85]

See merge request [!357]

  [#85]: gitlab.dune-project.org/NoneNone/issues/85
  [!357]: gitlab.dune-project.org/core/dune-istl/merge_requests/357


Closes #85

Note that this changes the default criterion to the symmetric one that
uses FirstDiagonal for computing strength of connection. Previously it
was the unsymmetric one with rowSum.

Note also that the new default will not work for complex.

Merge branch 'feature/novlap-scalar' into 'master'

ref:core/dune-istl This should make support for scalar matrices more complete.
Sequentially these matrices work, also for the overlapping case. It is just
the nonoverlapping that was missed.

@gruenich Your choice whether this is release material. Either PdeLab does not
use scalar matrices yet, or nobody tested the non-overlapping stuff with it.

@oliver.sander Can you check, whether I used the Hybrid stuff correctly? I
just looked at your changes to gsetc.hh as an example for this.

See merge request [!348]

  [!348]: gitlab.dune-project.org/core/dune-istl/merge_requests/348

This should make support for scalar matrices more complete.