Fix input arguments to dolfinx.mesh.refine_plaza

[francesco-ballarin]

Reported at https://fenicsproject.discourse.group/t/input-for-mesh-refinement-with-refine-plaza/13426

It seems to me that the current python API is certainly wrong (the option argument is not used anywhere), but I can't understand how it is possible that tests on main are currently passing since:

• a required input argument to the nanobind wrapped function is not being passed in from python, but nanobind is not complaining.
• the only two tests that exercise refine_plaza were providing input arguments in the wrong order.

Opening this as draft because I have no experience with mesh refinement, so I'd need help from someone else if the fix is more involved than sorting out wrong order of input arguments.

[francesco-ballarin]

Either I misinterpreted the role of the second input argument to refine_plaza (most likely, given my lack of experience in this field)

Indeed, I did misinterpret it ;) I was passing facets in, rather than edges.

I pushed the additional test and I am marking this as ready for review.

[jorgensd]

The issue with the old code was that it actually didn't send in edges (which is a kwarg), and the other kwargs weren't given by name, hence applying them to the wrong variable.

Redistribute became the option, and thus the tests passed.
The fix in the PR seems good.