[KV Sharing] Raise error if using eagle with fast prefill

[sarckk]

Purpose

Raise error if EAGLE spec decoding used with fast prefill. EAGLE expects logits for all tokens to be correct after prefill, this is violated by fast prefill.

Also consolidate warnings about fast prefill in one place.

Test Plan

vllm serve meta-llama/Llama-3.1-8B-Instruct --speculative-config='{"method": "eagle", "model": "yuhuili/EAGLE-LLaMA3.1-Instruct-8B", "num_speculative_tokens": 3}' -tp 1 --kv-sharing-fast-prefill

Test Result

(APIServer pid=1041573)   File "/data/users/yhshin/gitrepos/vllm/vllm/config/__init__.py", line 3676, in __post_init__
(APIServer pid=1041573)     raise NotImplementedError(
(APIServer pid=1041573) NotImplementedError: Fast prefill optimization for KV sharing is not compatible with EAGLE as EAGLE requires correct logits for all tokens while fast prefill gives incorrect logits for prompt tokens.

---

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.

[gemini-code-assist]

Code Review

This pull request correctly raises a NotImplementedError when EAGLE-style speculative decoding is used with fast prefill, addressing an incompatibility between the two features. The implementation is sound, with a clear error message. The change also refactors related checks and warnings into a more suitable location, improving code organization. The changes are well-executed and I have no suggestions for improvement.