feat(launch): re-add llm-d multinode support to infx.launch

[adibarra]

Summary

Re-adds llm-d (llmd-vllm) multinode support on the Python launcher from #3576, which dropped the launch_gb200-nv.sh llm-d branch.

llm-d keeps its own Slurm orchestration (submit.sh → sbatch job.slurm → srun + pyxis per node → server.sh). The launcher owns everything around it:

• policy.launch_path: multinode framework == llmd-vllm → LaunchPath.LLMD, enabled on gb200-nv only (LLMD_CLUSTERS); the table check requires slurm.squash.
• drivers/llmd.py: resolves the checkpoint from runners.yaml, imports the image via backend.prepare_image, runs llm-d/submit.sh directly with the point's topology, sequence lengths and concurrency, attaches to the printed job id, streams the log, checks the final Slurm state, and stages result JSONs, agentic and eval artifacts plus the server-log bundle. No per-model wrapper script.
• srt/models.py: dsv4 / fp4 / llmd-vllm on gb200-nv → DeepSeek-V4-Pro@numa1, served as deepseek-ai/DeepSeek-V4-Pro.
• dsv4-fp4-gb200-llmd-vllm: re-adds the low-latency 8k1k point (1P DEP8 + 1D TP8, conc 1) from the key removed in [Klaud Cold] Enact the September 8, 2026 DeepSeek-V4-Pro Single-turn 8k1k deprecation / 执行 2026 年 9 月 8 日 DeepSeek-V4-Pro 单轮 8k1k 场景下线 #2921, to exercise the path.

Fix: successful jobs no longer end CANCELLED

job.slurm used to scancel its own allocation once the decode coordinator finished, so every successful run ended CANCELLED 0:0, which the launcher reports as a failure (run 36720223235 finished its benchmark and was still marked failed). job.slurm now stops the srun step and exits 0, so the job ends COMPLETED. A step that dies before the coordinator finishes still fails the job with its exit code.

Driver port by @ilmarkov from #2719.

Testing

• pytest infx/tests/launch infx/tests/clusters infx/tests/matrix infx/tests/workflows: 1230 passed; ruff clean
• validate_perf_changelog against origin/main: passes
• Local harness for the job.slurm step control: hung workers + done marker → exit 0; step failing first → its exit code
• GB200 e2e: https://github.com/SemiAnalysisAI/InferenceX/actions/runs/36746690132

[github-actions]

Thanks for the contribution!

• Review: If this PR changes files owned by someone other than a repository admin or @SemiAnalysisAI/core, ask one eligible CODEOWNER to complete the latest PR_REVIEW_CHECKLIST.md before contacting a core maintainer on Slack. Follow the template exactly, including As a PR reviewer and CODEOWNER, I have reviewed this and have, so sign-off verification triggers.
• PR verification: Sweeps only run on labeled PRs. Add full-sweep-fail-fast (strongly recommended); use full-sweep-enabled only when matrix jobs should continue after a failure.
• After merging: PR authors must ensure all GitHub Actions jobs pass. Transient failures often pass on rerun; see how to rerun failed jobs.

中文

感谢你的贡献！

• **审阅：**如果 PR 修改的文件归属于仓库管理员及 @SemiAnalysisAI/core 之外的 CODEOWNER，请先联系一位有资格的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，再通过 Slack 联系核心维护者。必须严格遵循模板，并保留 As a PR reviewer and CODEOWNER, I have reviewed this and have，才能触发签核验证。
• **PR 验证：**扫描仅在带有标签的 PR 上运行。强烈建议添加 full-sweep-fail-fast；仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。
• **合并后：**PR 作者必须确保所有 GitHub Actions 任务通过。临时性失败通常可以通过重新运行恢复；参见重新运行失败任务的说明。