[Feature] Fix mixed cache-aware

[mouxinqq]

修复mixed cache-aware策略中的释放和selectworker逻辑

[paddle-bot]

Thanks for your contribution!

[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.

✅ mouxinqq
❌ mouxin

---

mouxin 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.}

[PaddlePaddle-bot]

🤖 AI Code Review | 2026-04-01

📋 Review 摘要

PR 概述：修复 mixed 模式下 cache-aware 调度策略中的 SelectWorker 和资源释放逻辑
变更范围：golang_router/internal/gateway/completions.go
影响面 Tag：Scheduler APIServer

问题

未发现阻塞性问题。

总体评价

本次变更正确修复了非 PD（mixed）模式下 cache-aware 策略的两个问题：

1. SelectWorker 调用修复：原代码传递空字符串 ""，导致 cache-aware 调度策略无法根据请求内容做出正确决策。现在正确提取 message 并传递给 SelectWorker。

2. ReleasePrefillTokens 调用补充：在 defer 释放资源时新增 ReleasePrefillTokens 调用，确保 token 计数器正确递减，与 PD 模式（Splitwise）的实现保持一致。

代码实现清晰，与已有的 PD 模式逻辑对齐，建议合入。