fix(config): annotate DSV4 TRT AgentX native KV offload

[cquil11]

Summary

• mark the six DeepSeek-V4 TensorRT-LLM AgentX points added by config(dsv4): add GB300 Dynamo-TRT AgentX points / 添加 GB300 Dynamo-TRT AgentX 配置点 #2690 as DRAM KV offload
• record the native backend version from the TensorRT-LLM image
• set the generated CPU DRAM budget to 193 GB, matching the recipes pinned to a 180 GiB host cache

Scope

This changes master-config metadata only. It does not modify benchmark recipes or runtime behavior, so there is no performance changelog entry.

Validation

• generated dsv4-fp4-gb300-dynamo-trt-agentx successfully
• confirmed all six points emit kv-offloading=dram, native@1.3.0rc24, and total-cpu-dram-gb=193
• git diff --check

---

Note

Low Risk
Master-config metadata only; no recipe, runtime, or benchmark-path changes.

Overview
Marks the six DeepSeek-V4 TensorRT-LLM AgentX search points as DRAM KV offload with the native backend (native@1.3.0rc24) and sets dram-utilization so generated CPU DRAM budget is 193 GB.

Metadata-only change in nvidia-master.yaml; recipes and runtime behavior are unchanged.

^{Reviewed by Cursor Bugbot for commit 6404c1b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[github-actions]

Thanks for the contribution! Please reach out to respective companies' CODEOWNER to fill in the latest PR_REVIEW_CHECKLIST.md before pinging core maintainer on Slack for review. In order for the signoff PR check bot to trigger, you must follow the PR_REVIEW_CHECKLIST.md template correctly, including the phrase As a PR reviewer and CODEOWNER, I have reviewed this and have.

For PR verification, add the full-sweep-fail-fast label (strongly recommended) to this PR — the benchmark sweep only runs on labeled PRs. Use full-sweep-enabled only if you need matrix jobs to keep running past a failure.

PR authors are responsible for ensuring that after merging, all GitHub Action jobs fully pass. A lot of the time, failures are just flakes and simply re-running the failed jobs will fix it. See GitHub's docs on re-running failed jobs

---

感谢你的贡献！请联系相应公司的 CODEOWNER 填写最新的 PR_REVIEW_CHECKLIST.md，然后再在 Slack 上联系核心维护者进行审阅。为了触发 signoff PR 检查机器人，你必须正确遵循 PR_REVIEW_CHECKLIST.md 模板，包括保留英文语句 As a PR reviewer and CODEOWNER, I have reviewed this and have。

如需进行 PR 验证，请为此 PR 添加 full-sweep-fail-fast 标签（强烈推荐）— 基准测试 sweep 仅在带有标签的 PR 上运行。仅当需要矩阵任务在失败后继续运行时才使用 full-sweep-enabled。

PR 作者有责任确保合并后所有 GitHub Action 任务完全通过。 很多时候失败只是偶发抖动（flake），重新运行失败的任务即可解决。参见 GitHub 关于重新运行失败任务的文档