dflash2 lane: drop drafter B12X (kv_cache_dtype constraint; main keeps B12X)

recipes/glm53-flash-nvfp4-r17abi-dflash2/recipe.toml

@@ -142,7 +142,7 @@ args = [
     # SPECULATIVE config, not the global flag (eagle/utils.py:
     # _make_eagle_draft_vllm_config) — B12X here mirrors the main model's.
     "--speculative-config", """
-'{"method":"draft_model","model":"incoai/GLM-5.3-Flash-DFlash2","num_speculative_tokens":7,"attention_backend":"B12X"}'""",
+'{"method":"draft_model","model":"incoai/GLM-5.3-Flash-DFlash2","num_speculative_tokens":7}'""",
     "--language-model-only",
     # MTP + CUDA graphs degrades OUTPUT QUALITY on the glm53 day-0 dev build
     # (forum #381350/#44, CONFIRMED by A/B bench 2026-08-28: same seed/flags,