Improvement: Use the directly calculated nodal saturation in the pressure coupling term

[wenqing]

In the current source code, the nodal saturations for the repressure coupling term use the extrapolated nodal values, which are inaccuracy in the local assembly. In this PR, such nodal saturations are calculated directly from the nodal capillary pressures. The reference results of the associated benchmarks have been updated (ufz/ogs5-benchmarks_ref#35).

[norihiro-w]

looks good

[wenqing]

@norihiro-w This change causes a problem for the nodes that shared by elements with different materials. That is from element to element, a node gets different saturation. Before PR#138, that nodal saturation is an extrapolated value and it is unique. I will drop this commit.

[norihiro-w]

@wenqing I'm unsure if it make sense to revert the commit. Though i don't check the details, unless the saturation is not a primary variable at nodes, the variable can be discontinuous across element. Here the nodal saturation are calculated only because we need integration point values.

[wenqing]

@norihiro-w The better way is to include the saturation during the integration of the pressure coupling matrix. The current implementation calculates the integration of the pressure coupling matrix without saturation for some general purpose, and then the nodal pressures are multiplied with nodal saturation. If nodal saturation is not unique, some problems with multi-material zones may have non-linear convergence problem.

[norihiro-w]

@wenqing i see. we should evaluate the saturation during the integration in the future.