Replace std::result_of by std::invoke_result_t

See merge request !85

The former is deprecated, and not supported by all compilers anymore.

[cleanup] Remove deprecated header arithmetic.hh

See merge request !84

This has been deprecated and superseded by `Dune::MatrixVector::`
long time ago (about 8 years).

[cmake] Update build system

See merge request !82

The creation of the library needs to be updated, because
otherwise the library is not found by downstream modules.

[CI][bugfix] Fix CI config

See merge request !83

* Use 2.10 image for 2.10 gcc job
* Remove 2.10 clang job, because there's no recent enough compiler
* Stop using outdated `dune:git-*` images

Replace std::result_of_t by std::invoke_result_t

See merge request !81

std::result_of_t is deprecated in C++17, and removed in C++20:

  https://en.cppreference.com/w/cpp/types/result_of

Minor fixes for finding IPOpt

See merge request !80

Starting with IPOpt 3.14 there is an overload of the Infeasibilities
method with five arguments. The old one with four arguments still
exists, but using it triggers a deprecation warning.

Use the new method if the IPOpt version is high enough.

This makes the code slightly longer, but also easier to understand.

At least newer Debian installations seem to used that path.

This fixes a CMake warning.

Set version to 2.11-git

See merge request !78

2.9 and 2.10 are what is currently in Debian stable and testing,
respectively.

Deprecate the file tuplevector.hh

See merge request !77

This was using a matrices as rows which should
never have worked. Surprisingly it only failed
with the recent modifications on how `field_type`
is computed in dune-istl.

2.9 is the latest version available as binary packages.
I like to be able to work with those as much as possible.

Replace the old ones by what dune-fufem uses currently.

This brings us std::tuple_element for Dune::Solvers::TupleVector.

An equivalent implementation of TupleVector has existed in dune-common
since 2016.  Please use that from now on.

Allow to set matrix hierarchy manually in MultiGridStep

See merge request !75

UMFPackSolver for MultiTypeBlockMatrix

See merge request !74

This is helpful when using AMG, where the matrix hierarchy
is computed while coarsening. Another application are
problems, where the coarse matrix is a preconditioner
that does not coincide with the Galerkin-restriction
of the fine matrix, e.g. for DG methods.

Remove 'blocksize' member from MultigridStep

See merge request !72

It seems to have no functionality and is incompatible with MultiTypeBlockVector

Stop using deprecated type PQkLocalFiniteElementCache

See merge request !71

Do not hardwire BitsetVector as default in IterationStep

See merge request !57

Use DefaultBitVector_t instead.  That way, the default has to be
overridden less frequently.

CHANGELOG.md: Latest change belongs to 2.10

See merge request !70

Implement generic proximal Newton solver for solving nonsmooth minimization problems

See merge request !63