[Misc] Upgrade vllm hash to 12_14

[Potabk]

What this PR does / why we need it?

Does this PR introduce any user-facing change?

1. fix [v1] Add PrefixLM support to FlexAttention backend vllm#27938
2. fix [Model][6/N] Improve all pooling task | Support chunked prefill with ALL pooling vllm#27145
   pooling models now supports chunked prefill and prefix caching,
3. fix [Model] Move multimodal_cpu_fields definition to field config vllm#30181
   define the CPU fields in the field config where they really belong.
4. fix [Core][MM] Add mechanism to configure multimodal fields which should stay on CPU vllm#28168
   define the CPU fields in the field config where they really belong.
5. fix kv_transfer: Rename the shared storage connectors vllm#30201
   some moudle rename
6. fix [MoE][Refactor] Make select_experts a non-static method vllm#29067
   fusedmoe moudle refactor
7. fix [MoE][Refactor] Remove most arguments to FusedMoEMethodBase.apply vllm#29066
   fusedmoe moudle refactor
8. fix [Attention] Make seq_lens_cpu optional in CommonAttentionMetadata to enable true async spec-decode vllm#29624

How was this patch tested?

• vLLM version: v0.12.0
• vLLM main: vllm-project/vllm@ad32e3e

[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]

Code Review

This pull request primarily focuses on upgrading the vLLM hash and ensuring compatibility with the new version. The changes are well-structured, involving API adaptations, refactoring for backward compatibility (e.g., expert_map handling), and introducing version-conditional logic, particularly for vLLM v0.12.0. Overall, the changes appear correct and necessary for the upgrade. I've identified one minor issue in a test file where a model is duplicated, which should be addressed to avoid redundant test runs.

[github-actions]

This pull request has conflicts, please resolve those before we can evaluate the pull request.

[Potabk]

follow up:1. qwen3-next refactor; 2. npu_model_runner get_attn_backend remove

[Potabk]

the full CI passed here https://github.com/vllm-project/vllm-ascend/actions/runs/20224367586/job/58052479148?pr=5000