add oob_tcp_if_include for OMPI and NCCL_SOCKET_IFNAME parameters for RCCL to multinode

[paklui]

Motivation

add oob_tcp_if_include for OMPI and define NCCL_SOCKET_IFNAME parameters for RCCL to multinode

Technical Details

I tried to run earlier on a couple of MI350 systems for multi-node RCCL, run into this issue that the default socket for RCCL and OMPI does not work.
Since there is already oob_port specified for btl_tcp_if_include, could we also use it for oob_tcp_if_include for OMPI and NCCL_SOCKET_IFNAME that I could work around this issue.

The error I get is the following if I don't use the socket.
The test would hang for awhile and time out with the following:

node-h30-08:395097:395195 [6] NCCL INFO RAS client listening socket at ::1<28028>
[2025-12-04 19:41:31] node-h28-08:1921104:1921160 [4] /longer_pathname_so_that_rpms_can_support_packaging_the_debug_info_for_all_os_profiles/src/out/rhel-9.5/9.5/build/rccl/hipify/src/misc/socket.cc:589 NCCL WARN socketPollConnect poll() returned 1, no POLLOUT events
node-h28-08:1921104:1921160 [4] NCCL INFO /longer_pathname_so_that_rpms_can_support_packaging_the_debug_info_for_all_os_profiles/src/out/rhel-9.5/9.5/build/rccl/hipify/src/misc/socket.cc:641 -> 2
node-h28-08:1921104:1921160 [4] NCCL INFO /longer_pathname_so_that_rpms_can_support_packaging_the_debug_info_for_all_os_profiles/src/out/rhel-9

Test Plan

with this change, I am able to execute all combination of rccl multinode tests using:

source ./test_venv/bin/activate
pytest -vvv --log-file=/tmp/test.log -s ./tests/rccl/rccl_multinode_cvs.py --cluster_file input/cluster_file/cluster.json  --config_file input/config_file/rccl/rccl_config.json --html=./rccl.html --capture=tee-sys --self-contained-html

Test Result

The RCCL multinode test pass. Almost all pass using the latest git.
ufw is a separate issue since it does not work on CentOS 9 (default OS for GT 1.5 systems)

---------------------------------------- Generated html report: file:///apps/paklui/cv350/cvs/rccl.html ----------------------------------------
=============================================================== short test summary info ================================================================
FAILED tests/rccl/rccl_multinode_cvs.py::test_disable_firewall - Failed: Following FAILURES seen - ['Service ufw not disabled properly on node node-h30-08', 'Service ufw not disabled properly on node node-h28-08']
====================================================== 1 failed, 47 passed in 1219.50s (0:20:19) =======================================================

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.