Allow Mfront behavior to work with PK2 stress

[bleyerj]

Currently we impose MFront behaviors to work with PK1/F total Lagrangian formulation, we should also allow to change this behavior to work in PK2/EGL at least (Eulerian formulation with Cauchy seems less useful)
Pinning @nagelt for reference