Improve pipeline for voxel meshes

[chr-engwer]

You are using hex-meshes mainly to represent voxel meshes.

In this case you should be able to use some improvements:

1. it might be better to use the sub-entity center search instead of the normal projection, as you have a structured (voxel) mesh
2. instead of writing an unstructured grid file for a structured grid it would be more appropriate to go for the structured grid and the filter cells using dune-subgrid, like we do it for node shift meshes (where we also start with the voxel mesh).

[juangpc]

What implications would this have in the future? I would prefer (from a limited understanding) the unstructured grid options even if it requires more memory/computation-time. When dealing with mri->mesh libraries we cannot limit our work to those libraries that generate a structured grid. Or perhaps I'n not really understanding the point?

[chr-engwer]

You should distinguish between

• unstructured tet meshes
• voxel meshes
• voxel meshes with node-shift and potentially in the future
• unstructured hex meshes

The unstructured hex mesh is currently used, but it imposes more costs. I don't suggest to remove this code path. The proposal is to add an improved code path for voxel meshes.

[juangpc]

Thanks.