fix(critical): fold max_completion_tokens + max_num_seqs=2 + max_model_len=80000 + xllm_latest layer import

Sub 655 root causes (confirmed from log analysis):
1. protocol.py: max_completion_tokens never folded into max_tokens
   → 162/881 replay requests rejected 400 (extra_forbidden)
2. max_num_seqs=1 → t2_n_2 test fails (needs n=2)
3. max_model_len=131072 → OOM crash at 62% replay, opencompass all 0

Fixes:
- protocol.py: model_validator fold_max_completion_tokens
- yaml: max_num_seqs=2, max_model_len=80000, PYTORCH_CUDA_ALLOC_CONF
- topk_softmax stays =0 (corex CUB BlockReduce incompatible on BI-V100)

xllm_latest import to ex_engine/:
- npu_torch layers: GDN(1164L), Qwen3.5 GDN, attention, fused_moe
- cuda/moe kernels: topk_softmax_kernels.cuh, moe_combine, moe_compute_index
- npu kernels: causal_conv1d, recurrent_gated_delta_rule
- model headers: qwen3_5.h, qwen3_next.h

ex_engine/xllm_models/llm/qwen3_5_mtp.h

@@ -0,0 +1,59 @@
+/* Copyright 2025-2026 The xLLM Authors.
+
+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 <memory>
+
+#include "models/llm/qwen3_5.h"
+#include "models/llm/qwen3_5_mtp_base.h"
+#include "models/model_registry.h"
+
+namespace xllm {
+
+class Qwen3_5MtpModelImpl final : public Qwen3_5MtpModelImplBase {
+ public:
+  explicit Qwen3_5MtpModelImpl(const ModelContext& context)
+      : Qwen3_5MtpModelImplBase(context) {}
+};
+
+class Qwen3_5MtpForCausalLMImpl final : public Qwen3_5MtpForCausalLMImplBase {
+ public:
+  explicit Qwen3_5MtpForCausalLMImpl(const ModelContext& context)
+      : Qwen3_5MtpForCausalLMImplBase(
+            context,
+            std::make_shared<Qwen3_5MtpModelImpl>(context)) {}
+};
+TORCH_MODULE(Qwen3_5MtpForCausalLM);
+
+REGISTER_CAUSAL_MODEL(qwen3_5_mtp, Qwen3_5MtpForCausalLM);
+REGISTER_CAUSAL_MODEL(qwen3_5_moe_mtp, Qwen3_5MtpForCausalLM);
+
+REGISTER_MODEL_ARGS_LOADER(qwen3_5_mtp,
+                           [](const JsonReader& json, ModelArgs* args) {
+                             return qwen3_5_mtp::load_model_args(
+                                 json, args, "qwen3_5_text", "qwen3_5_mtp");
+                           });
+
+REGISTER_MODEL_ARGS_LOADER(qwen3_5_moe_mtp,
+                           [](const JsonReader& json, ModelArgs* args) {
+                             return qwen3_5_mtp::load_model_args(
+                                 json,
+                                 args,
+                                 "qwen3_5_moe_text",
+                                 "qwen3_5_moe_mtp");
+                           });
+
+}  // namespace xllm