[box][compositional] Dirichlet condition for box method does not use setState method
When using the box-method the setState method which is used in compositional models to set the state of your primary variables is not used. It looks like the primary variables are always pressure and saturation. Therefore it is not possible to use a one-phase system and a mole fraction instead of saturation.
The (initial) solution can have different states than the Dirichlet values, either because they are differently set at the beginning or in time dependent Dirichlet BC problems (connection to #545 (closed), should be actually a problem for tpfa then as well). The residual is then wrong because it is set (sol - Dirichlet) and sol and dirichlet have different states.