Leith+E bugfix (#334)

Gustavo Marques noticed that with `LEITH_CK = 0.0` we were still
getting negative Laplacian viscosity coefficients southwest of
land. There's an if statement designed to limit the magnitude
of the backscatter so that it doesn't overcome the biharmonic
damping. When c_K = 0.0 *and* when smoothed vorticity at the
q point is zero, this if statement ironically causes the
backscatter coefficient to equal its largest possible value.
Smoothed vorticity at the q point is always zero on land, so
cells whose northwest corner are on land always have the max
backscatter coefficient (which then gets smoothed into the
interior). This commit zeros out the backscatter in those
cells.

src/parameterizations/lateral/MOM_hor_visc.F90

@@ -1339,6 +1339,7 @@ subroutine horizontal_viscosity(u, v, h, uh, vh, diffu, diffv, MEKE, VarMix, G,
               else
                 m_leithy(i,j) = CS%m_leithy_max(i,j)
               endif
+              m_leithy(i,j) = G%mask2dBu(i,j) * m_leithy(i,j)
             endif
           enddo ; enddo