tnl-benchmark-blas produces wrong log

The log file generated by tnl-benchmark-blas has wrong format in case of SpMV tests.

It happens only when the benchmark runs on CPU only.

It produces the following log:

: title = Sparse matrix-vector multiplication (double)
: precision = double
: loops = 10
: host name = yvonne
: architecture = x86_64
: system = Linux
: system release = 4.15.0-34-generic
: start time = Thu Dec 13 2018, 14:30:54
: number of MPI processes = 1
: OpenMP enabled = true
: OpenMP threads = 12
: CPU model name =  Intel(R) Core(TM) i7-5820K CPU @ 3.30GHz
: CPU cores = 6
: CPU threads per core = 2
: CPU max frequency (MHz) = 3600
: CPU cache sizes (L1d, L1i, L2, L3) (kiB) = 32, 32, 256, 15360

! rows
! columns
! elements per row
! Matrices::CSR
!! CPU
!!! bandwidth
!!! time
!!! speedup
100000
100000
5
    48.956
    0.000190234
    N/A

! rows
! columns
! elements per row
! Matrices::CSR
!! Matrices::Ellpack
!!! CPU
!!!! bandwidth
!!!! time
!!!! speedup
100000
100000
5
    46.8707
    0.000198698
    N/A

! rows
! columns
! elements per row
! Matrices::CSR
!! Matrices::Ellpack
!!! Matrices::SlicedEllpack
!!!! CPU
!!!!! bandwidth
!!!!! time
!!!!! speedup
100000
100000
5
    2.60511
    0.00357494
    N/A

! rows
! columns
! elements per row
! Matrices::CSR
!! Matrices::Ellpack
!!! Matrices::SlicedEllpack
!!!! Matrices::ChunkedEllpack
!!!!! CPU
!!!!!! bandwidth
!!!!!! time
!!!!!! speedup
100000
100000
5
    9.98237
    0.000932956
    N/A

! rows
! columns
! elements per row
! Matrices::CSR
!! Matrices::Ellpack
!!! Matrices::SlicedEllpack
!!!! Matrices::ChunkedEllpack
!!!!! Matrices::CSR
!!!!!! CPU
!!!!!!! bandwidth
!!!!!!! time
!!!!!!! speedup
200000
200000
5
    5.09024
    0.00365922
    N/A

! rows
! columns
! elements per row
! Matrices::CSR
!! Matrices::Ellpack
!!! Matrices::SlicedEllpack
!!!! Matrices::ChunkedEllpack
!!!!! Matrices::CSR
!!!!!! Matrices::Ellpack
!!!!!!! CPU
!!!!!!!! bandwidth
!!!!!!!! time
!!!!!!!! speedup
200000
200000
5
    16.4476
    0.00113246
    N/A

....

closed via commit e3c60cbf

mentioned in commit e3c60cbf