Add qwen235b prefill-only tuning support

configs/examples/dash0_qwen235b_prefill_thinking_baseline.json

@@ -0,0 +1,13 @@
+{
+  "observation": "This is the prefill-only baseline aligned to run_qwen235b.sh. Keep the internal vLLM launch shape unchanged and replay the thinking trace with completion forced to 1 token.",
+  "diagnosis": "A baseline measurement is required before proposing prefill-only TTFT improvements. Preserve all current envs and flags from run_qwen235b.sh to establish the first feasible sampling_u/request_rate point under the TTFT-only SLO.",
+  "config_patch": {
+    "env_patch": {},
+    "flag_patch": {}
+  },
+  "expected_effects": [
+    "Establish a launch-safe prefill-only baseline for qwen3-235b thinking traffic",
+    "Seed later trials from the first feasible sampling_u if one exists"
+  ],
+  "why_not_previous_failures": "No previous failures in this study."
+}