About Boussinesq's turbulent viscosity hypothesis: historical remarks and a direct evaluation of its validity