Bug explanation

As a pre-processing step of writing data to HDF5 files the mesh element types of the mesh are analyzed
If all elements of the mesh have the same element type this information can be used to write the topological data to the HDF5 in a more efficient way
In parallel simulations the sub-meshes are partitioned according to the partition of the bulk mesh
Each MPI process computes the element type information for its own sub-mesh
There can be cases where the sub-mesh consists of elements of the same type, but not every partition takes a part of the sub-mesh
As a consequence different MPI processes compute either the default element type on a empty sub-mesh part or the real element type on a non-empty sub-mesh part, so the information is contradicting each other
This leads to wrong calculation of offsets and finally invalid state

Feature description was added to the changelog - No new feature - only bug fix
Tests covering your feature were added?
Any new feature or behaviour change was documented? - Fixed bug / wrong behaviour

Fix XDMF/HDF5 output on sub-meshes