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/docs/mkdocs/overrides/.icons/email-fill.svg

@@ -0,0 +1 @@
+<?xml version="1.0" standalone="no"?><!DOCTYPE svg PUBLIC "-//W3C//DTD SVG 1.1//EN" "http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd"><svg t="1755326167427" class="icon" viewBox="0 0 1024 1024" version="1.1" xmlns="http://www.w3.org/2000/svg" p-id="13611" xmlns:xlink="http://www.w3.org/1999/xlink" width="256" height="256"><path d="M926.47619 355.644952V780.190476a73.142857 73.142857 0 0 1-73.142857 73.142857H170.666667a73.142857 73.142857 0 0 1-73.142857-73.142857V355.644952l304.103619 257.828572a170.666667 170.666667 0 0 0 220.745142 0L926.47619 355.644952zM853.333333 170.666667a74.044952 74.044952 0 0 1 26.087619 4.778666 72.704 72.704 0 0 1 30.622477 22.186667 73.508571 73.508571 0 0 1 10.678857 17.67619c3.169524 7.509333 5.12 15.652571 5.607619 24.210286L926.47619 243.809524v24.380952L559.469714 581.241905a73.142857 73.142857 0 0 1-91.306666 2.901333l-3.632762-2.925714L97.52381 268.190476v-24.380952a72.899048 72.899048 0 0 1 40.155428-65.292191A72.97219 72.97219 0 0 1 170.666667 170.666667h682.666666z" p-id="13612" data-spm-anchor-id="a313x.search_index.0.i13.7cb13a81l9Qfa6" class="selected" fill="#acacac"></path></svg>