feat: support multinode for json format output

[shadychan]

Support multinode when testing with json format output (-j), which fixed these issues:

• Missing device info of remote nodes in json output
• Unexpected local node d2d bandwidth value due to variable localDevice not being updated through MPI

$ mpirun -n 8 --npernode 4 -hostfile /opt/tiger/hostfile --mca orte_tmpdir_base /opt/tiger/openmpi_tmp --mca pml ob1 --mca btl ^openib,smcuda --mca btl_tcp_if_include eth0 --mca oob_tcp_if_include eth0 nvbandwidth -t multinode_device_to_device_memcpy_write_ce -j
{
        "nvbandwidth" : 
        {
                "CUDA Runtime Version" : 13000,
                "Driver Version" : "580.105.08",
                "GPU Device list" : 
                [
                        "0: NVIDIA GB200 (00000008:06:00): (n179-067-085)",   <== missing device info of remote nodes
                        "1: NVIDIA GB200 (00000009:06:00): (n179-067-085)",
                        "2: NVIDIA GB200 (00000018:06:00): (n179-067-085)",
                        "3: NVIDIA GB200 (00000019:06:00): (n179-067-085)"
                ],
                "git_version" : "v0.8",
                "testcases" : 
                [
                        {
                                "bandwidth_description" : "memcpy CE GPU(row) <- GPU(column) bandwidth (GB/s)",
                                "bandwidth_matrix" : 
                                [
                                        [
                                                "N/A",
                                                "3231.73",  <== unexpected local node d2d bandwidth value
                                                "3235.38",
                                                "3232.94",
                                                "775.055",
                                                "775.195",
                                                "775.125",
                                                "775.125"
                                        ],
                                        [
                                                "3232.94",
                                                "N/A",
                                                "3232.94",
                                                "3232.94",
                                                "775.055",
        ...
}

• MPI_ABORT invoked in multinode broadcast testcases and also in test with option -j -p multinode

$ mpirun -n 8 --npernode 4 -hostfile /opt/tiger/hostfile --mca orte_tmpdir_base /opt/tiger/openmpi_tmp --mca pml ob1 --mca btl '^openib,smcuda' --mca btl_tcp_if_include eth0 --mca oob_tcp_if_include eth0 nvbandwidth -t multinode_device_to_device_broadcast_one_to_all_sm -j
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 4 in communicator MPI_COMM_WORLD
with errorcode 1.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
[n179-067-191:06592] 5 more processes have sent help message help-mpi-api.txt / mpi-abort
[n179-067-191:06592] Set MCA parameter "orte_base_help_aggregate" to 0 to see all help / error messages

[shadychan]

@ramasubramanianrahul @deepakcu @esitaridi
Please review, thanks.