On the use of wall functions as boundary conditions for two-dimensional separated compressible flows