[Spec Decode] Support EAGLE for Qwen3 MoE

[seven-mile]

Purpose

Fix #25927 and #25134.

Test Plan

vllm serve \
    Qwen/Qwen3-30B-A3B \
    --host 0.0.0.0 \
    --port 7000 \
    --seed 42 \
    -dp 2 \
    --enable-expert-parallel \
    --enforce-eager \
    --max-model-len 4096 \
    --gpu_memory_utilization 0.8 \
    --speculative-config '{"model":"Tengyunw/qwen3_30b_moe_eagle3","num_speculative_tokens":4}'

lm_eval --model local-completions \
  --tasks gsm8k \
  --model_args model=Qwen/Qwen3-30B-A3B,base_url=http://0.0.0.0:7000/v1/completions,num_concurrent=1,max_retries=3,tokenized_requests=False

Test Result

w/ spec

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
gsm8k	3	flexible-extract	5	exact_match	↑	0.8484	±	0.0099
		strict-match	5	exact_match	↑	0.8984	±	0.0083

w/o spec

Tasks	Version	Filter	n-shot	Metric		Value		Stderr
gsm8k	3	flexible-extract	5	exact_match	↑	0.8506	±	0.0098
		strict-match	5	exact_match	↑	0.8984	±	0.0083

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @seven-mile.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[seven-mile]

Just forgot to mark it ready. Close since a similar patch #26485 has started review.