Interface solver results in potentially non-differentiable function

See the discussion in !3956 (merged): discussion. Brief summary:

• File: dumux/porousmediumflow/2p/saturationreconstruction.hh
• if (pc < pcMin && pcMin > 0.0) clamps to zero, which leads to potentially unsmooth function and degrading Newton performance
• We should find a C1 expression that we can always use, e.g. 1.0 -pcSw.sw(interfacePcSw.pc(1.0-sn))
• but we could of course also use another smooth regularization that handles sn<0 but smoothly transfers to 1.0 -pcSw.sw(interfacePcSw.pc(1.0-sn))

For the case interfacePcSw == pcSw this was solved in !3956 (merged) by simply using sn in this case (no reconstruction necessary). interfacePcSw == pcSw means that the control volume has the parameters that have the lowest capillary pressure at 100% wetting saturation of all laws present at the material interface.

added Discussion label

added Physics / Math label

added Porous Medium Flow label

changed the description

/cc @martins

assigned to @martins

unassigned @martins