[Feat] Add draft_logprobs for Speculative Decode MTP

[sunlei1024]

描述:
本 PR 为 MTP 的 Speculative Decode MTP 功能增加 draft_logprobs 支持，并在 OpenAI 兼容接口上做了对应扩展。

主要改动:

1. 新增请求参数 include_draft_logprobs

   • 在 /chat/completions 和 /completions 接口中，用户可以通过该参数控制是否返回 draft_logprobs。

2. 新增响应参数 draft_logprobs

   • 当 include_draft_logprobs 为 true 时，响应中会包含 draft_logprobs，用于记录推测解码过程中中间的概率值。

3. 改动说明: 非Speculative Decode MTP 核心解码逻辑保持不变，与现有解码流程兼容。

目的:

• 增强对 Speculative Decode 的可观测性和调试能力。
• 完善 OpenAI 兼容接口，方便用户在调试或实验中获取中间预测概率信息。

示例请求 (curl)：

1. /completions 接口

curl https://{ip}:{port}/v1/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "your-model",
    "prompt": "Hello, world!",
    "logprobs": 5,
    "include_draft_logprobs": true
  }'

2. /chat/completions 接口

curl https://{ip}:{port}/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "your-model",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello, world!"}
    ],
    "logprobs": true,
    "top_logprobs": 5,
    "include_draft_logprobs": true
  }'

示例响应片段：

{
  "id": "cmpl-xxx",
  "object": "text_completion",
  "choices": [
    {
      "text": "Hello",
      "logprobs": [ ... ], 
      "draft_logprobs": [ ... ]
    }
  ]
}

{
  "id": "chatcmpl-xxx",
  "object": "chat.completion",
  "choices": [
    {
      "message": {
        "role": "assistant",
        "content": "Hello",
        "logprobs": [ ... ], 
        "draft_logprobs": [ ... ]
      }
    }
  ]
}

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

✅ sunlei1024
❌ sunlei18

---

sunlei18 seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}