minimum_xspacing does not account for the immersed boundary

[simone-silvestri]

I think it's because of these definitions

Oceananigans.jl/src/ImmersedBoundaries/immersed_boundary_nodes.jl

Lines 67 to 69 in 504ab5d

	xspacings(ibg::ImmersedBoundaryGrid, ℓ...) = xspacings(ibg.underlying_grid, ℓ...)
	yspacings(ibg::ImmersedBoundaryGrid, ℓ...) = yspacings(ibg.underlying_grid, ℓ...)
	zspacings(ibg::ImmersedBoundaryGrid, ℓ...) = zspacings(ibg.underlying_grid, ℓ...)

This causes the CFL to be limited by possibly immersed nodes.
It is quite relevant on the TripolarGrid where nodes near the north poles are immersed with the spacings tending to zero.

I can open a PR to fix this.

[glwagner]

Nice find. Also we should note that the immersed boundary grid can change the spacings (in addition to the fact that we must omit immersed spacings). For example with partial cells the zspacings are different than on the underlying grid.

[glwagner]

cc @ali-ramadhan I'm not sure if this was bundled with other changes you made but a good thing to be aware of