[2p][adaptive] use right privars for vol vars update

For cell-centered scheme, the primary variables in the reconstruction were not divided by the number of children that composed the primary variables in a father element. This does not have an effect here as the helper class and the tests only consider incompressible fluids. Furthermore, the problem would only appears if in one refinement some elements get refined more than one level. However, for the sake of readability it should be correct here.