* Do not pass base class as template parameter for compositional
  localRes and fluxVars

* instead pass base class as template parameter for rans localRes
  and fluxVars

Thanks to Kilian for contributing the actual change of all the test's spatial params!

What was broken:
The test used philtophob law which used the wrong residual saturations. It went into efftoabslaw with
the non-wetting saturation so swr and snr where the wrong way around. It also overwrote the krw/krn function
of the material law in use with a quadratic/cubic law. It also defined the capillary pressure negative but then
it would be pc = pw-pn.

What was fixed:
Use the wettingPhase function instead to make the gas phase the wetting phase

This changed:
- the pw solution as swr and snr were corrected (the wrong way around pw was equal)
- the pc solution because it is now positite
- the mobilities as they didn't use the real van genuchten curves

* add test
* add PDELab solution for comparison
* speed-up by avoid calling getparam
* improve determination of the gradient
* increase verbosity for comparison

Fluid systems exported phase indices for wetting and nonwetting components. The fluid system
is not the right place to decide about wettability. This introduces wettingPhase() in the spatial
params which returns the index of the phase in the fluidsystem we consider the wetting phase.
This way wettability can change temporally and spatially.
Indices throughout the 2p/2p2c/2pnc nc and mpnc where renamed accordingly.

Dennis Gläser authored 6 years ago and Timo Koch committed 6 years ago

Contributions by Timo Koch, Kilian Weishaupt.