Correct regularization for ImmersedBoundaryCondition

[simone-silvestri]

I am implementing immersed drag for a SeaIceModel here, but I noticed that the regularization for ImmersedBoundaryCondition is not correct because it assigns boundary conditions in nothing directions, for example, in CliMA/ClimaSeaIce.jl#84 this happens:

julia> grid = RectilinearGrid(size = (10, 10, 1), x = (0, 1), y = (0, 1), z = (0, 1));

julia> grid = ImmersedBoundaryGrid(grid, GridFittedBottom(rand(10, 10)));

julia> u_bcs = FieldBoundaryConditions(grid, (Face, Center, Nothing); immersed = FluxBoundaryCondition(1));

julia> model = SeaIceModel(grid; boundary_conditions = (; u = u_bcs));

julia> model.velocities.u.boundary_conditions.immersed
ImmersedBoundaryCondition:
├── west: Nothing
├── east: Nothing
├── south: FluxBoundaryCondition: 1.0
├── north: FluxBoundaryCondition: 1.0
├── bottom: FluxBoundaryCondition: 1.0
└── top: FluxBoundaryCondition: 1.0

In this case, top and bottom boundary conditions should be a Nothing, so this PR corrects this regularization

[navidcy]

just noting that even after this is merged; using ClimaSeaIce#main and Oceananigans#main I get:

(ClimaSeaIce) pkg> st
Project ClimaSeaIce v0.3.2
Status `~/Library/CloudStorage/OneDrive-TheUniversityofMelbourne/Documents/Research/ClimaSeaIce.jl/Project.toml`
  [79e6a3ab] Adapt v4.3.0
  [63c18a36] KernelAbstractions v0.9.38
  [9e8cae18] Oceananigans v0.97.3 `https://github.com/CliMA/Oceananigans.jl.git#main`
  [7181ea78] RootSolvers v0.4.4
  [f2b01f46] Roots v2.2.8
  [d496a93d] SeawaterPolynomials v0.3.10

julia> using Oceananigans, ClimaSeaIce

julia> grid = RectilinearGrid(size = (10, 10, 1), x = (0, 1), y = (0, 1), z = (0, 1));

julia> grid = ImmersedBoundaryGrid(grid, GridFittedBottom(rand(10, 10)));

julia> u_bcs = FieldBoundaryConditions(grid, (Face, Center, Nothing); immersed = FluxBoundaryCondition(1));

julia> model = SeaIceModel(grid; boundary_conditions = (; u = u_bcs));

julia> model.velocities.u.boundary_conditions.immersed
ImmersedBoundaryCondition:
├── west: Nothing
├── east: Nothing
├── south: FluxBoundaryCondition: 1.0
├── north: FluxBoundaryCondition: 1.0
├── bottom: FluxBoundaryCondition: 1.0
└── top: FluxBoundaryCondition: 1.0

[navidcy]

I think this happens because Oceananigans.Operators.assumed_field_location(:u) = (Face, Center, Center) and Oceananigans.BoundaryConditions.regularize_field_boundary_conditions use that, even for the SeaIceModel

[glwagner]

good call. Then perhaps SeaIceModel needs to do something special with the bcs