[Refactor] Add expert processed token count output for DispatchFFNCombine/DispatchFFNCombineBF16

[guanguan0308]

What this PR does / why we need it?

Add New Output for Expert Token Count
An additional output tensor expert_token_nums is added to both operators to meet the requirement of tracking token distribution among experts:

Tensor Name: expert_token_nums
Dimension: 1D tensor
Shape: (local_expert_num,)
Data Type: int32
Semantics: Represents the number of tokens actually received by each expert on the current card.

Does this PR introduce any user-facing change?

How was this patch tested?

• vLLM version: v0.14.1
• vLLM main: vllm-project/vllm@dc917cc

[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 introduces a new output, expert_token_nums, to the DispatchFFNCombine operator and its BF16 variant. The changes correctly propagate this new output through the operator's definition, host-side functions, and kernel implementations. However, there are a few issues to address. Firstly, two header files contain redundant inclusions, which should be cleaned up for better code quality. More importantly, adding a new required output is a breaking API change. It appears that downstream consumers, such as the Python bindings and tests, have not been updated to reflect this change, which will likely cause them to fail.

[github-actions]

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