data: port complete MoE + xllm layer call chains from upstream repos

MoE call chain from ds_vllm (vllm-project/vllm latest):
  ex_engine/moe/ — 20 files, 8736 lines
  - modular_kernel.py (1630 lines) — base classes for modular MoE
  - experts/fused_batched_moe.py (972 lines) — NaiveBatchedExperts
  - prepare_finalize/batched.py (171 lines) — token grouping by expert
  - topk_weight_and_reduce.py (176 lines) — scatter-add finalize
  - fused_moe.py (1740 lines) — main fused_moe dispatch
  - config.py (1407 lines) — FusedMoEQuantConfig
  - activation.py, utils.py, layer.py, etc.

xllm layer code (jd-opensource/xllm):
  ex_engine/xllm_layers/ — 39 files, 5859 lines
  - ilu/fused_moe.cpp (797 lines) — production ixformer 7-step MoE pipeline
  - ilu/attention.cpp (189 lines) — paged_attention + flash_attn bridge
  - npu_torch/qwen3_gated_delta_net_base.cpp (576 lines) — GDN reference
  - common/rms_norm.cpp, rotary_embedding.cpp, activation.cpp, dense_mlp.cpp

xllm ILU kernels — synced 10 files to upstream (diffs from prior edits)

These are reference implementations, NOT hand-written.
Source repos: vllm-project/vllm, jd-opensource/xllm

ex_engine/xllm_layers/common/dense_mlp.h

@@ -0,0 +1,67 @@
+/* Copyright 2025 The xLLM Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    https://github.com/jd-opensource/xllm/blob/main/LICENSE
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#pragma once
+
+#include <torch/torch.h>
+
+#include "activation.h"
+#include "framework/model/model_args.h"
+#include "framework/parallel_state/parallel_args.h"
+#include "framework/quant_args.h"
+#include "framework/state_dict/state_dict.h"
+#include "linear.h"
+
+namespace xllm {
+namespace layer {
+
+class DenseMLPImpl : public torch::nn::Module {
+ public:
+  DenseMLPImpl() = default;
+  DenseMLPImpl(int64_t hidden_size,
+               int64_t intermediate_size,
+               bool is_gated,
+               bool has_bias,
+               const std::string& hidden_act,
+               bool enable_result_reduction,
+               const QuantArgs& quant_args,
+               ProcessGroup* process_group,
+               const torch::TensorOptions& options,
+               const std::string& module_prefix = "");
+
+  torch::Tensor forward(const torch::Tensor& hidden_states);
+
+  void load_state_dict(const StateDict& state_dict);
+  void load_state_dict(const StateDict& state_dict,
+                       const std::vector<std::string>& gate_up_name,
+                       const std::string& down_name);
+
+  // Get FP8 input scale from gate_up_proj for fused RMSNorm+FP8 quantization
+  std::optional<torch::Tensor> get_fp8_input_scale() const;
+
+ private:
+  bool is_gated_;
+  int64_t intermediate_size_;
+  ProcessGroup* process_group_;
+  ColumnParallelLinear gate_up_proj_{nullptr};
+  RowParallelLinear down_proj_{nullptr};
+  Activation act_{nullptr};
+  bool is_smoothquant_;
+  std::string hidden_act_;
+};
+TORCH_MODULE(DenseMLP);
+
+}  // namespace layer
+}  // namespace xllm