Bug in temp_mld and salt_mld diagnostics

[rmholmes]

I think there is a bug in the mixed layer depth integrated tracer diagnostics, temp_mld and salt_mld, with spurious zero values found in some continental shelf regions. This is visible in the second panel below from an ACCESS-OM2-025 simulation averaged over a month, where the panel is computed via temp_mld/mld/rho0.

I don't know where the error is coming from. I have an alternative diagnostic in my own branch which fixes this as part of a project on mixed layer tracer budgets that I will submit a PR on at some point.

I doubt anyone is using these diagnostics but posting just in case...