SimAI1.5 update

[MXtremist]

• Advanced Inference Simulation: Model complex scenarios with Prefill/Decode separation.
• Modern Model Support: Now includes DeepSeek, Qwen3Moe and Qwen3Next.
• Request Scheduling: Request scheduling is now handled by a component adapted from Microsoft's Vidur.

[tianhao909]

Good Job！ Well Done！ I have two minor review comments for this pr. Thanks～

a. 【Run Llama-3-8B's three branches】We can remind users to download the relevant dataset from /vidur/data (https://github.com/microsoft/vidur/tree/main/data) and copy it to the /vidur-alibabacloud/data directory (this instruction can be added to the README).

b. 【Run DeepSeek-671B with AICB】To ensure successful reproduction, we may need to provide a few sample workload files from the current /aicb/results/workload directory (pending confirmation on whether these workloads can be open-sourced). This would be particularly helpful for users who do not have access to a cluster environment to run AICB's DPSK and Qwen3.

=========
a. 【Run Llama-3-8B的三个分支】我们可以提醒用户补充下载相关数据集 /vidur/data （https://github.com/microsoft/vidur/tree/main/data），复制粘贴到/vidur-alibabacloud/data位置 （补充到readme就可以）
b. 【Run DeepSeek-671B with AICB】如果要跑通，可能需要我们提供几个目前的/aicb/results/workload示例 （不确定这些workload示例能不能开源～） （如果用户没有环境集群运行aicb的dpsk和qwen3的话）