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/xllm/.gitmodules

@@ -0,0 +1,57 @@
+[submodule "third_party/brpc"]
+	path = third_party/brpc
+	url = https://gitcode.com/xLLM-AI/brpc.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/cpprestsdk"]
+	path = third_party/cpprestsdk
+	url = https://gitcode.com/xLLM-AI/cpprestsdk.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/minja"]
+	path = third_party/minja
+	url = https://gitcode.com/xLLM-AI/minja.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/sentencepiece"]
+	path = third_party/sentencepiece
+	url = https://gitcode.com/xLLM-AI/sentencepiece.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/smhasher"]
+	path = third_party/smhasher
+	url = https://gitcode.com/xLLM-AI/smhasher.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/xllm_ops"]
+	path = third_party/xllm_ops
+	url = https://gitcode.com/xLLM-AI/xllm_ops.git
+	fetchRecurseSubmodules = true
+[submodule "third_party/etcd_cpp_apiv3"]
+	path = third_party/etcd_cpp_apiv3
+	url = https://gitcode.com/xLLM-AI/etcd-cpp-apiv3.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/spdlog"]
+	path = third_party/spdlog
+	url = https://gitcode.com/xLLM-AI/spdlog.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/Mooncake"]
+	path = third_party/Mooncake
+	url = https://gitcode.com/xLLM-AI/Mooncake.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/torch_npu_ops"]
+	path = third_party/torch_npu_ops
+	url = https://gitcode.com/xLLM-AI/torch_npu_ops.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/cutlass"]
+	path = third_party/cutlass
+	url = https://gitcode.com/xLLM-AI/cutlass.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/xllm_atb_layers"]
+	path = third_party/xllm_atb_layers
+	url = https://gitcode.com/xLLM-AI/xllm_atb_layers.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/xxHash"]
+	path = third_party/xxHash
+	url = https://gitcode.com/xLLM-AI/xxHash.git
+	fetchRecurseSubmodules = false
+[submodule "third_party/tilelang-ascend"]
+	path = third_party/tilelang-ascend
+	url = https://gitcode.com/xLLM-AI/tilelang-ascend.git
+	branch = ascendc_pto
+	fetchRecurseSubmodules = true