feat(CRITICAL): ix_bridge call chain + upstream xllm/ds_vllm sync

3 changes that close the MoE performance gap:

1. _custom_ops.py: Add ix_bridge as Priority 0 for topk_softmax
   - Before: tries our .cu kernel (fails) → PyTorch fallback (1-3 TPS)
   - After: tries ix_bridge → ixformer::infer::topk_softmax() → FAST
   - Call chain: _custom_ops.topk_softmax() → ix_bridge.topk_softmax()
     → ix_moe_bridge.so → ixformer::infer::topk_softmax()

2. Dockerfile: Add ix_moe_bridge.cpp precompile step
   - This was the missing link: code existed but was never compiled
   - Uses torch.utils.cpp_extension.load() to link against libixformer.so

3. upstream_ref sync from GitHub (cloned, not rewritten):
   - xLLM-AI/xllm: ILU kernels + CUDA MoE + GDN fp32 state mgmt
   - Deep-Spark/vllm: latest MoE kernel sources

Dockerfile

@@ -26,7 +26,13 @@ RUN chmod +x /workspace/qwen3_6_scripts/patch_ops.sh && \
     bash /workspace/qwen3_6_scripts/patch_ops.sh 2>&1 | tee /workspace/patch_ops.log ; \
     echo "[Dockerfile] patch_ops exit code: $?"
 
-# Step 5: Precompile GDN kernel (needs vllm in path, so after patch_ops)
+# Step 5: Precompile ix_moe_bridge.cpp → links to ixformer::infer::topk_softmax()
+# This is the C++ pybind bridge that makes ixformer SDK callable from Python.
+# Source: upstream_ref/xllm/xllm/core/kernels/ilu/fused_moe.cpp call pattern
+RUN python3 /workspace/ex_engine/precompile_ix_bridge.py 2>&1 | tee -a /workspace/ex_build.log ; \
+    echo "[Dockerfile] ix_bridge precompile exit code: $?"
+
+# Step 6: Precompile GDN kernel (needs vllm in path, so after patch_ops)
 RUN python3 /workspace/qwen3_6_scripts/precompile_gdn.py \
     /workspace/qwen3_6_scripts/flash_qla_sm70 2>&1 | tee -a /workspace/ex_build.log ; \
     echo "[Dockerfile] gdn precompile exit code: $?"