Decouple physics and discretization (Do not always include all discretization schemes for flux laws?)
When making a change, e.g., to the MPFA-version of Fick's law, all porousmediumflow and freeflow test have to be rebuild.
Is there a way to circumvent this?
This goes in a similar direction as #583.
Do we really need to include the actual implementations in the flux headers? Couldn't we include them somewhere else?