Updated the computation of the Bernulli beams.

Before the computation of mass of a beam did not consider the cross section area.
It assumed that StructuralMaterial::rho had a unit of "mass per unit length.
This commit changes this and now the density is handled with its real units.

The old behaviour can be restored by defined the macro AKANTU_STRUCTURAL_MECHANICS_OLD_MASS_COMPUTATION.

It is important that all test are passed.
The reason for this is, that they usually set the corss section to $1$.
And thus the units of StructuralMaterial::rho does not has an influence on the result.