1) When we expect a high oscillation of the flow residuals and, especially,
of the objective value, it is recommended to use averaging (see corresponding
dict in the primal solver setup). Even with this engineering trick, optimisation
in cases involving practically unsteady of highly oscillating flows, using
steady-state adjoint solutions should be used with caution.
2) The adjoint equations, with the current setup, have an almost monotonic
convergence. Hence, no averaging is needed since the last time-step will be the
more converged one.
