GitLab

* added new CI jobs for parallel execution and reference tests
* added Muphi reference test which now reads tolerance explicitly
* Fixed computation of L2 error and default tolerances in compare_grid

We trust dune-randomfield now

* reduce number and runtime of ODE tests
* use layered parfield from dorie-pub benchmarks
* choose tight residual error limits
* disable all other tests for now

Also add path to dunecontrol as variable in gitlab-ci.yml

Meta-ini references the asset path

- unset DUNE_BUILD_TESTS_ON_MAKE_ALL
- add test-mass-conservation as regular executable
- remove EXCLUDE_FROM_ALL property of from test targets explicitly

Test executable is now run with new metaini file.
Limits are set closely to the current mass conservation results.
Moved class definitions into own header file.
Test executable is managed from within the `testing` directory

This fixes #39.

This effectively reverts 9c168d22

removed ode tests with adaptation for now. increased adapt threshold for ref tests to account for new estimator

This reverts commit 27a3d22d.

added tests for comparing sequential and parallel results. slightly altered reference.py for this task