- the develop branch requires an autoconf version newer than may be
  available on the user's system

- Also optimizations in creation of coherent mesh -- domainDecomposition::parallelMesh()
- Also optimizations in polyMesh::write() when in coherent format and the associated work in construction of SlicePermutation.

- unifying subset identification for processor patch construction in CoherentMesh
- Offsets as iterable of pairs
- Offsets construction with globalIndex
- move processor patch communication into CoherentMesh as member functions
- separating permutations into SlicePermutation and FragmentPermutation (SRP)
- add sliceWritePrimitives for const label* data
- removing and refactoring redundant functionality

- replace constexpr-if constructs
- replace std::make_unique
- replace generic lamdas
- add trailing return type with automatic return types
- simplified is_range type trait

- allows selection of different system compilers.
  Eg,

     export WM_COMPILER=Gcc
     export WM_COMPILE_CONTROL="version=11"

  will use gcc-11 and g++-11 for compilation

ENH: naive decomposition on-the-fly if the size of partitionStarts does not fit to the number of MPI ranks, i.e. if not present at all
BUG: proper checkOut of boundary_ in polyMesh

- COHERENT: worse formatting but readable.

mesh and fields are stored in data.bp and ascii metadata files
-> consistent layout in time folders and constant/polyMesh
- propagate pathname through all ADIOS2 functionality
- removing redundant usages, includes, and dependencies to wrapping functions to ADIOS2
- adding initialization strategy for FieldComponents
- minor formatting

- Non-uniform internal field in coherent format: Communicate fields on
  processor patches from lower ranks to higher ones since these are
  only present once in a coherent surface field.
- Handle non-proc patches separately

- Use writeFormat entry from controlDict for decision while reading,
  since no format mixing is allowed, e.g. writing in coherent and
  reading in ascii is forbidden.