Inferred vertical diffusivity calculation

[jbisits]

When calculating the vertical component of the buoyancy gradient there are many zeros (unsurprisingly as the two layers have uniform S and T). This means that when using Integral for the volume integral the result is NaN.

[jbisits]

To solve this I have set the computation to be zero if ∂z_b = 0 to avoid getting Inf (after trying to divide by zero) and then NaN when sum! acts for the Integral.