avoid tuncation error through integer division and promote to use "bigint"...

  if (i == 3) return 1.0*allbonds/atom->natoms;

  if (i == 4) {

    int nlocal = atom->nlocal;

    int nbonds = 0;   // BIGINT?

    const int nlocal = atom->nlocal;

    bigint nbonds = 0;

    for (int j = 0; j < nlocal; j++)

      nbonds += numbond[j];

    int allbonds;

    MPI_Allreduce(&nbonds,&allbonds,1,MPI_INT,MPI_SUM,world);

    int allneigh;   // BIGINT?

    MPI_Allreduce(&list->ipage->ndatum,&allneigh,1,MPI_INT,MPI_SUM,world);

    double neighsperatom = allneigh/atom->natoms; 

    double bondsperatom = 0.5*allbonds/atom->natoms;

    bigint allbonds;

    MPI_Allreduce(&nbonds,&allbonds,1,MPI_LMP_BIGINT,MPI_SUM,world);

    bigint allneigh,thisneigh;

    thisneigh = list->ipage->ndatum;

    MPI_Allreduce(&thisneigh,&allneigh,1,MPI_LMP_BIGINT,MPI_SUM,world);

    const double natoms = atom->natoms;

    const double neighsperatom = static_cast<double>(allneigh)/natoms;

    const double bondsperatom = 0.5*static_cast<double>(allbonds)/natoms;

    return neighsperatom * bondsperatom;

  }