Sediment models will not work with hydrostatic RK3 time stepping

[ali-ramadhan]

I noticed that https://github.com/OceanBioME/OceanBioME.jl/blob/main/src/Models/Sediments/coupled_timesteppers.jl uses type piracy to couple a sediment model with an Oceananigans model for time stepping.

I'm wondering if there's a cleaner solution here. Maybe a simulation callback? It's more work on the user to add the callback, but then there won't be a need for type piracy and to maintain separate time_step! functions.

I'm not sure if we can use hydrostatic RK3 time-stepping (CliMA/Oceananigans.jl#3930) with OceanBioME.jl because of this, but it looks like it's just isolated to using the sediment models?

[glwagner]

yikes!

[jagoosw]

Sediment time-stepping definitely needs to be overhauled!

We could probably do it in update_biogeochemical_state! like how particles are stepped