Pseudo dissipation with tuple closures

[tomchor]

At the moment we can't get (pseudo)-dissipation when the model has a tuple closure.

I'm not sure if it's possible to get actual dissipation when the diffusivity isn't isotropic, but it should be straightforward to get pseudo-dissipation.

[glwagner]

We also can't get dissipation when the closure doesn't use an eddy diffusivity, right? So it's not possible to calculate diffusivity for any turbulence closure.

[tomchor]

We also can't get dissipation when the closure doesn't use an eddy diffusivity, right? So it's not possible to calculate diffusivity for any turbulence closure.

True. Although we could consider throwing a useful error in those cases?

I think we should at least support the common cases (like (VerticalDiffusivity(), HorizontalDiffusivity())) though, especially because it should be pretty straightforward.

[glwagner]

Agree, it makes sense to try to support any AbstractScalarDiffusivity since it's calculable and there's an interface. We can also support tuples of AbstractScalarDiffusivity I think, via https://github.com/CliMA/Oceananigans.jl/blob/ce5ba71449cf16b8737dca10997a6119dee4c665/src/TurbulenceClosures/abstract_scalar_diffusivity_closure.jl#L61-L62