Bug#1094984: python-meshplex: FTBFS with numpy 2.x

[Drew Parsons]

Package: python3-meshplex
Followup-For: Bug #1094984
Control: tags -1 ftbfs
Control: block -1 by 1094985

The bug is effectively in meshio, meshplex is simply accessing a mesh
returned by meshio.

The problem is that the 2D cell type ("triangle") has no dtype
attached to the cell array.  Evidently in numpy 2 this defaults to
dtype float, even if the values are integers.

By constant the 3D cell type ("tetrahedron") extracted from the same
mesh comes with dtype=int64.

In meshio source where get_cells_type is defined,
a default dtype=int is used for empty cell lists. I think the failing
meshplex test is using this for tetra. The test mesh is 2D, and the
test checks if the list is empty.

I'm not where exactly meshio is setting up the cell list (with no
explicit dtype).  numpy 2 bug reports have already been filed against
meshio (Bug#1094985, https://github.com/nschloe/meshio/issues/1499 )