Fix few bugs + improve docstring of SplitExplicitFreeSurface

[navidcy]

bugs:

• Δy = minimum_xspacing(grid)
• some leftover Δt_max was changed to max_Δt
• when computing substeps via cfl then we need to ensure it's an integer

[glwagner]

Might be better to contribute this to the broader user API overhaul in #3124 (won't this cause conflicts?)

[navidcy]

#3124 was merged!

[navidcy]

@simone-silvestri, what exactly is max_Δt? It's the maximum baroclinic tilmestep, right?

[glwagner]

@simone-silvestri, what exactly is max_Δt? It's the maximum baroclinic tilmestep, right?

That's right. Maybe we could have a better name for that.

[navidcy]

Well I at least added a remark.

[navidcy]

Perhaps at

Oceananigans.jl/src/Models/HydrostaticFreeSurfaceModels/split_explicit_free_surface.jl

Line 280 in f273c54

substeps = Int(round(2 * max_Δt / Δtᴮ))

we should use Int(floor(...)) to be tad more conservative.

[simone-silvestri]

Perhaps at

Oceananigans.jl/src/Models/HydrostaticFreeSurfaceModels/split_explicit_free_surface.jl

Line 280 in f273c54

substeps = Int(round(2 * max_Δt / Δtᴮ))

we should use Int(floor(...)) to be tad more conservative.

probably ceil(Int, ...)