chore(recipes): delete unreferenced srt-slurm recipes

[adibarra]

Summary

First PR in the recipe refactor stack. Deletes srt-slurm recipe files and override blocks that nothing references: no master config row, launcher table (including the power.py recipe globs), test, or doc.

Removed	Count	Why it is dead
minimaxm3/vllm/gb200-fp4/agentx/{agg-dep4,agg-dep4-vllm-simple,agg-dep8,agg-tp4,agg-tp4-vllm-simple,disagg-1p1d-dep8-dep4}.yaml	6	Superseded by the bundles in the same directory
qwen3.5/sglang/gb200-fp4/agentx/agg-tp4-mtp-hicache.yaml	1	Superseded by agg-variants.yaml:override_tp4
single_node/.../minimaxm3/vllm/mi355x-fp4-mtp/agentic.yaml	1	No master row uses it
override_mtp_tp4, zip_override_mtp_lowlat, override_mtp_maxtpt_7p2d, override_mtp_maxtpt_4p1d in dsr1/sglang/b200-fp4/8k1k/disagg-stp-mtp-variants.yaml	4 blocks	Master rows select only the STP variants; the MTP points use their own per-topology files

Kept on purpose, although unreferenced:

• The 35 recipes under the retired */1k1k/ directories, as a historical record.
• configs/dsv4-moe-load-balancer-configs/*, which RECIPES.md documents keeping.

Verification

• python -m infx.matrix.generate full-sweep over both master configs, with and without --all-evals, is byte-identical before and after (1767 and 1797 entries, including node-count and recipe-fingerprint).
• Each referenced selector of the trimmed bundle (zip_override_stp_lowlat[0..2], override_stp_tp4, override_stp_maxtpt_7p2d) resolves to the same recipe before and after through selected_recipes.
• pytest infx/tests/matrix infx/tests/srt_slurm infx/tests/launch: 732 passed.

No perf-changelog entry: nothing that runs changes.

[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 run only when the PR appends an inferencex-e2e/perf-changelog.yaml entry and carries exactly one primary label: full-sweep-fail-fast (strongly recommended; canary plus per-matrix fail-fast), full-sweep-enabled (canary; matrix jobs continue after a failure), or non-canary-full-sweep-enabled (no canary or fail-fast). The modifiers all-evals, evals-only, and agentx-fast require a primary label. On fork PRs, a maintainer applies the label. See sweep labels and reuse.
• 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：**只有当 PR 在 inferencex-e2e/perf-changelog.yaml 末尾追加了条目，并且恰好带有一个主标签时，才会运行扫描：full-sweep-fail-fast（强烈推荐；canary 加逐矩阵 fail-fast）、full-sweep-enabled（有 canary；矩阵任务在失败后继续运行）或 non-canary-full-sweep-enabled（无 canary，也无 fail-fast）。修饰标签 all-evals、evals-only 和 agentx-fast 必须与主标签一起使用。fork PR 的标签由维护者添加。参见扫描标签与复用。
• **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。