Introduce epsilon for floating point comparison in test.

This commit contains backward imcompatible. All flux variables
now have to be default constructed. Implementers of flux variables
need to make their implementation default construcable too.
All calls to the old non-default ctors will throw errors at runtime and
emit deprecation warning at compile time. FluxVars are no constructed
and update like this:

FluxVariables fluxVars;
fluxVars.update(...)

Being default contructable offers more flexible class inheritance
and more flexible caching possiblities.

To this end especially the 2pnc/2pncmin flux variables where
updated to comform with other models. This involves some
deprecations. The 2pncmin reference had to be altered in the
velocity. The velocity was previously outputted wrongly as
the correct volumeFlux variable was never computed for 2pnc/2pncmin.

Ignore the porosity for all box models since it is defined element-wise
in these test but the default 2p2c implementation outputs porosity per
vertex. Depending on the order of the elements, the porosity would
differ in these cases.

Add Property for maximum number of subdomains
Fix Richardson number
Fix Indentation

A base problem on the model level was necessary

Timo Koch authored 9 years ago and Christoph Grüninger committed 9 years ago

Using correct floating point comparison with epsilons led
to a alightly different solution. Reference updated.

Use input gridcreator instead of structured dgf file

Get rid of external grid managers
Use coarser mesh (65x65 instead of 85x85)

Argument dependent lookup deduces the Dune namespace from
the function arguments that are also in namespace Dune.

only the ni-base-model will be listed in the module list
ni-test-problems include docu from non-ni-base problem

Implement a temperature() function in the volvars and use it
instead of volvars.fluidstate().xtemperature().
This reuqired the implementation of two template functions
to ensure that the correct temperature() function is called
depending on the number of energy equations considered.

Use volVars.moleFraction(wPhaseIdx, NaClIdx) instead
of volVars.fluidState().moleFraction(wPhaseIdx, NaClIdx)

Implement a pressure()  and density() function in the volvars and use it
instead of volvars.fluidstate().x(phaseIdx).

After commit 8557b77e
which fixed shadowed variables in the fluidsystem,
the test did not pass anymore. In addition, a new grid
creator is used now which also leads to slight differences in the
coordinates. Therefore, the reference solution had to be updated.

Fixes a bug with conditional jump over uninitialized values, which was intoduced by generalizing how massFractions are set.
This keeps the generalization, but reverts everything else to what it has been before.
Changes in reference solutions are necessary due to small changes.

Thanks gruenich for git-bisecting.

This changes the reference solutions.

This requires updating the reference solutions.
For mpnckinetic it was not able to check, whether the reference solution was already broken before.

The bounding box tree uses the grid's index set as it is only concerned
with elements no matter which discretization. We have to therefore use
the gridview's indexset to access the point source map in the problem.

* Implement problem interface functions mimicking the normal source interface
* Implement a possibility to implement derived point source classes getting cutomized lambda functions
	* the pointsource class gets an update method
        * An example implementation with test for time dependent point sources is implemented
* Restructure the point source class. Do not derive from PrimaryVariables (less felxible) instead overload
  some operators for convenience
* Call the solDependentPointSource function for each point source not per dof (multiple ps possible)
* Recompute the map for adapted grids

* Remove unused types
* Remove superfluous semicolons after methods