GM+E implementation inconsistent with paper

[iangrooms]

@sdbachman's GM+E parameterization is described here. The backscatter is accomplished by adding a negative viscosity that acts only on the barotropic component of the flow. In Scott's paper, the backscatter coefficient is proportional to minus the GM coefficient, but the proportionality is spatially varying and depends on the GM work rate and the barotropic lateral shear. See, e.g., equation (54) in Scott's paper.

The 'GME' parameterization in MOM6 is quite different from the GM+E parameterization from Scott's paper.
The commit message for this commit says that 'GME is now set to be some multiple of the GM coeff.'

The implementation in MOM6 has a barotropic backscatter, and the backscatter coefficient is proportinal to the GM coefficient. This line gets the GM coefficient. Next these lines use the GM coefficient times the efficiency factor to create a barotropic stress that then gets added to the existing 3D stress here. The problem is that the efficiency factor (computed here) doesn't depend on the GM work rate or the barotropic shear, so this version of GME is really not GM+E at all.