Inconsistent non-conservative mortars

[amrueda]

The non-conservative mortar fluxes for P4estMesh{2}, T8codeMesh{2}, TreeMesh{2}, and TreeMesh{3} are currently computed on the small faces in the outward-pointing direction of the small elements only, neglecting the non-conservative (asymmetric) nature of these fluxes.

For example, the non-conservative calc_mortar_flux! for P4estMesh{2} (see here) computes a single non-conservative flux as follows:

noncons = nonconservative_flux(u_ll, u_rr, normal_direction, equations)

A unique flux per mortar interface (flux + 0.5 * noncons) is then projected/copied to the neighbor elements on both sides of the mortar by mortar_fluxes_to_elements! here.

A fix for the 3D p4est solver is done in #2127.