Non cartesian cases for simple diffusion and the soret effect

[rekomodo]

#64

[RemDelaporteMathurin]

• Spherical cases can only be 1D
• It could be worth adding a few details about the derivation in non-cartesian (mostly the expression of the divergence)
• I think the MMS source isn't valid for the spherical case. I obtained $S = -6 D$ in the spherical system of coordinates
• In the write up the exact solution is $1 + r^2$ but not in the code block
• In cylindrical/spherical one cannot set a BC other than no flux on $r=0$ boundary because that would be a singularity. I think that in your case specifically, since the gradient is zero on this boundary it works well. But I would have the domain start at $r=1$.

[rekomodo]

It could be worth adding a few details about the derivation in non-cartesian (mostly the expression of the divergence)

The same expression from the implementation right? I'll add it at the end.

[rekomodo]

just rebased the branch