[CI] Main2main 0513

[Potabk]

What this PR does / why we need it?

1. fix [Bugfix] Fix SP pass for multimodal models and PP+SP residual handling vllm#33322
   overwrite gpu_modelrunner.sync_and_gather_intermediate_tensors, for the sceniro pp+sp+tp, skip scatter the residual for ascend

2. [Model Runner V2] support qwen35 / mamba hybrid model vllm#35520
   Adapted to the modifications of ModelRunner v2 for hybrid attn in interface level, .
   Todo: Added support for Mamba in ModelRunner in Ascend. any pull_request is welcome

3. [Aiter][ROCm] gdn_linear_attn kernel fusion vllm#40711

4. [Attention][Cleanup] Remove tree attention vllm#42121

5. [Model] use AutoWeightsLoader for DeepSeekV2 vllm#41706

6. [Core] Replace routing replay with device cache and async D2H pipeline vllm#39917
   Disable async_schedule when enable_return_routed_experts=True

7. [MoE Refactor] Move expert map related code into ExpertMapManager class vllm#41046

8. [MoE Refactor] EPLB refactoring for FusedMoE vllm#41055

9. [Model Runner V2] Apply synthetic mode to probabilistic rejection sampler vllm#41035

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.20.1
• vLLM main: vllm-project/vllm@c7aa186

[github-actions]

👋 Hi! Thank you for contributing to the vLLM Ascend project. The following points will speed up your PR merge:‌‌

• A PR should do only one thing, smaller PRs enable faster reviews.
• Every PR should include unit tests and end-to-end tests ‌to ensure it works and is not broken by other future PRs.
• Write the commit message by fulfilling the PR description to help reviewer and future developers understand.

If CI fails, you can run linting and testing checks locally according Contributing and Testing.

[gemini-code-assist]

Warning

Gemini encountered an error creating the summary. You can try again by commenting /gemini summary.