ref(upstream): FULL TREE — Deep-Spark xllm (1470) + ds_vllm csrc/models (703)

Replaces cherry-picked upstream_ref with complete source trees.

xllm/ — Iluvatar official C++ inference engine (15MB, 1470 files)
  Complete: kernels → layers → models → runtime → scheduler → api
  Excluded: .git, binary images, third_party submodule checkouts

ds_vllm/ — Iluvatar official vllm fork (8MB, 703 files)
  Included: csrc/ (ALL CUDA kernels), fused_moe/, qwen3_5 model, _custom_ops
  Excluded: tests, benchmarks, docs, examples (not needed for reference)

Critical call chains now fully traceable:
  MoE: moe_topk_softmax_kernels.cuh → ixformer.h → fused_moe.cpp → layer
  GDN: qwen3_gated_delta_net_base.cpp → qwen3_5_gated_delta_net.cpp
  Attention: ixformer.h → xllm_paged_attention → attention.cpp

upstream_ref/ds_vllm/csrc/torch_utils.h

@@ -0,0 +1,17 @@
+#pragma once
+
+// Shared TORCH_UTILS_CHECK across both libtorch stable and unstable source
+// files. Keep this header free of CUTLASS/CUTE so attention/quant headers can
+// use it.
+//
+// If TORCH_TARGET_VERSION is defined, we are building _C_stable_libtorch.so so
+// use STD_TORCH_CHECK via header-only.
+// Otherwise, use TORCH_CHECK via torch/all.h.
+
+#ifdef TORCH_TARGET_VERSION
+  #include <torch/headeronly/util/Exception.h>
+  #define TORCH_UTILS_CHECK STD_TORCH_CHECK
+#else
+  #include <torch/all.h>
+  #define TORCH_UTILS_CHECK TORCH_CHECK
+#endif