Add new testcase for porous convection model

I was just thinking that it would be good to put the initial disturbance on one side of the domain and see if we buoyancy driven cavity flow, with its cyclic flow around the perimeter of the square domain. With the initial disturbance in the center, we are relying on an instability to form the two opposing cycles of flow. But if we put the disturbance on the left, or we heat the domain from the opposite corners (hot in the bottom left, cold in the top right) we should see a stable cycle effect.

That stable cycle effect should help with V&V because we can get reliable solutions that converge to a stable solution as we decrease the mesh size.

_Originally posted by @jpfairbanks in https://github.com/AlgebraicJulia/Decapodes.jl/issues/297#issuecomment-2712257474_