- when combining lists in processor order this simplifies code and
  reduces memory overhead.

  Write this:
    ----
    labelList collected;

    const globalIndex sizing(input.size());
    sizing.gather(input, collected);
    ----

  OR

    ----
    labelList collected;
    globalIndex::gatherOp(input, collected);
    ----

  Instead of this:

    ----
    labelList collected;

    List<labelList> scratch(Pstream::nProcs());
    scratch[Pstream::myProcNo()] = input;
    Pstream::gatherList(scratch);

    if (Pstream::master())
    {
        collected = ListListOps::combine<labelList>
        (
            scratch,
            accessOp<labelList>()
        );
    }
    scratch.clear();
    ----