Initialization of OMIP simulation

[glwagner]

• Assess the differences between initializing from ERA5, NCAR SPEAR, ECCO, and WOA
• Utility for in-painting NaNs / land regions
• Potentially implement a utility for smoothing
• Declare victory when we can run a simulation from rest with prescribed temperature and salinity and no surface fluxes
• Potentially run a simulation with prescribed fluxes (eg not using JRA55 forcing) from ECCO

[glwagner]

@simone-silvestri I may need more consultation on methods for inpainting.

[glwagner]

Here's one idea: for inpainting, we can use a pure-ocean setup (with no continents), zero PrescribedVelocityField, constant horizontal diffusivity, convective adjustment + background vertical diffusivity, and strong relaxation applied to the points where initialized data exists. Then we will run the simulation for some time until the solution is smooth. This will diffusive the known data into land regions.