nested sparse vector

See merge request duneuro/duneuro!82

Unify drivers

See merge request duneuro/duneuro!66

- deleted old meeg folder and adapted CMakeLists.txt accordingly
- added featureManager analogously to before, instead of in the
meeg_driver_factory_impl it is now created in the
volume_conductor_factory_impl
- checks of configs are either carried out in the driver_interface (if
mutual) or in the respective (un)fitted_volume_conductor class

- for the parallel loop using tbb, the grainSize was read from the
config with a default of 16 in the unfitted case, while in the fitted
case no value was set; now, the unfitted approach is used for both

- grainSize for tbb used to be read from config with 16 being the
default in the unfitted case, in the fitted case it was not passed
(default from tbb), now the unfitted case is used for both

- avoids repetition of code in (un)fitted volume_conductor classes
- solveEEGForward_impl is a protected function in the base
volume_conductor_interface that is called in both derived classes in the
(un)fitted_volume_conductor with their specific solvers etc.
- "subtract_mean" was and still is only implemented in the fitted case
(the unfitted case requires a different implementation)

- see
https://gitlab.dune-project.org/duneuro/duneuro/wikis/code-formatting

the general idea is to seperate the volume conductor (fitted, unfitted)
from the driver in order to avoid many repetitions of code
(unfitted_meeg_driver, fitted_meeg_driver, unfitted_tdcs_driver,
fitted_tdcs_driver,...)

sparse vector did not represent the hierarchical structure of
the function space, which makes it difficult to properly map them
to a flat representation.

We now create sparse vectors also as a hierarchical structure and add the
necessary tags to handle them in the PDELab ISTL backend.

(currently debug output is disabled)

Resolve "Add manual grain size to applyTransfer in FittedMEEGDriver"

Closes #37

See merge request duneuro/duneuro!64

multipolar venant

See merge request duneuro/duneuro!80

Similar to vertex_based_venant_source_model.hh, calls
multipolar_venant.hh

MEG caching turned off by default

See merge request duneuro/duneuro!78

- as MEG caching is quite memory demanding, the default is set to false

Feature/2.7 cleanup

See merge request duneuro/duneuro!70

fix thread-safety when applying the transfer matrix

Closes #48

See merge request duneuro/duneuro!77

- due to the new feature management, the config was changed to be
non-const for the applyE/MEGTransfer methods in the drivers
- within the TBB block, the non-const version of ParameterTree::sub
modifies the config
- this can lead to unpredictable segfaults when applying the transfer
matrix in case TBB is enabled

Make P1GradientLocalFiniteElementMap compatible with PDELab 2.6 interfac changes

See merge request duneuro/duneuro!73

make use of GMP conditional

Closes #46

See merge request duneuro/duneuro!74

- replace GMPField usage by double if HAVE_GMP is false
- avoids compilation error when compiling duneuro (HAVE_UDG=1) without
the libgmp-dev package

Feature management

Closes #38 and #43

See merge request duneuro/duneuro!68

- Nüßing2016 instead of Nüßing2014 (led to error when print_citations
was called for UDG)
- DUNEuro instead of duneuro
- added missing dois/urns if existing
- consistent names in references (initials followed by last name)

dimension declared until now. Not downwards compatible.

- static member function in FeatureManager class to remove experimental
flag from type/solverType string in case it exists
- changed error message in source model factories and driver factory