[fix][richards] Extended richards: Fix balancequationoptions and do porosity update before evaluating effective diffusion coefficients

What this MR does / why does DuMux need it: The richards model never overwrote the balancequation options, which is why, although diffusion was calculated (for the extendend Richards model that considers water diffusion in the gas phase), it was immediatly also substracted, leading to a net diffusive flux of 0. With the richards-specific balancequationopts it depends on the phase index if diffusion is balanced or not. Additionally, the porosity in the volumevariables was updated at the wrong place, now it is updated before the effective diffusion coefficient. This correction also made it necessary to change the boundary condition in the richards evaporation tests. a fixed value does not work if the liquid phase actually disappears.

fixes #1134 (closed)