feat(amd): move AMD multi-node fixed-sequence benchmarks to srt-slurm

[cquil11]

Moves AMD multi-node fixed-sequence benchmarks to srt-slurm and removes the legacy bash that coordinated them. Based on #3352, which moves single-node fixed-sequence.

• Cluster setup hooks: per-allocation host checks live in runners/srt-slurm/hooks/<cluster>/setup.sh, run through the profile's default_host_setup, with shared helpers in hooks/common.sh. MI355X gets its RDMA/QoS preflight, GPU drain gate, and hugepage reclaim; MI300X checks MEC firmware (177 or newer) instead of setting container environment in a preamble. Documented in AGENTS.md and docs/configuration-procedures.md.
• runners/launch_mi355x-amds.sh: multi-node jobs with a CONFIG_FILE submit the recipe through apply_srt_recipe with the shared post-eval (eval runs drop fake expert dispatch), then collect results with copy_fixed_sequence_results / copy_agentic_results and eval artifacts from logs/eval_results. The legacy multi-node path now only serves AgentX.
• runners/srt-slurm/mi355x-amds.yaml: adds the multi-node settings (fabric interface, output dir, model path, HF cache mounts, exporter and nginx settings) to refactor: migrate fixed-sequence recipes to SRT-Slurm / 将定长配方迁移至 SRT-Slurm #3352's profile.
• Recipes for Qwen3.5 FP8/MXFP4 and DeepSeek-R1 FP8/MXFP4 disaggregated serving; the seven configs now select them through CONFIG_FILE. Recipes write results in the shared sa-bench layout.
• Adds sglang-disagg and vllm-disagg to synthetic acceptance.
• Removes the legacy fixed-sequence launch scripts, the amd_utils vLLM and ATOM engine paths, the fixed-sequence bench client, and the per-model fixed-sequence server settings.

Recipes own serving and workload settings; InferenceX does not rewrite them. Workflow settings reach custom benchmarks through the exported job environment.

AgentX configs (DeepSeek V4 DSpark, GLM-5.3 TileRT) stay on the legacy path until the AgentX migration.

Folds in #2628, #2629, #2631, #2633.