[Optimization] Reduce preemption occurrence when blocks not enough

[rainyfly]

Motivation

当解码 block不足时，调度会抢占正在解码的请求，释放对应的 block 资源。并分配给剩余的解码请求。
之前的调度逻辑，在 waiting 队列里有请求的时候，发现剩余的 block可以容纳下一条新请求部分的 chunk（new_token_num)，就会将其调度回去做 prefill。在 block已经严重不足的时候，会造成反复 抢占->调度 prefill->抢占->调度 prefill的重调度行为，造成性能下降。
为了解决这一问题，在调度新请求做 prefill 时，考虑给正在解码的请求预留部分 block，只有在给正在解码的 每条请求所预留的 block剔除后，并且剩余的 block 还可以容纳整条当前需要 prefill 的请求，才把请求从 waiting 队列里调度出来做 prefill。

Modifications

新增环境变量：FD_RESERVE_OUTPUT_BLOCK_NUM_FOR_DECODE_WHEN_SCHEDULE_NEW_PREFILL
含义：从 waiting 队列里调度新请求做 prefill 时，需要给每条正在解码的请求预留的 block 数量，默认为 16

Usage or Command

None

Accuracy Tests

None

Checklist

• Add at least a tag in the PR title.
  • Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
  • You can add new tags based on the PR content, but the semantics must be clear.
• Format your code, run pre-commit before commit.
• Add unit tests. Please write the reason in this PR if no unit tests.
• Provide accuracy results.
• If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

[paddle-bot]

Thanks for your contribution!

[Copilot]

Pull request overview

This PR optimizes the scheduler to reduce excessive preemption when GPU blocks are insufficient. The optimization introduces a reservation mechanism that reserves blocks for currently decoding requests before scheduling new prefill requests, preventing the scheduler from repeatedly preempting and rescheduling requests.

Key changes:

• Adds configurable environment variable FD_RESERVE_OUTPUT_BLOCK_NUM_FOR_DECODE_WHEN_SCHEDULE_NEW_PREFILL (default: 16 blocks per decoding request)
• Modifies scheduling logic to check if sufficient blocks remain after reserving blocks for running requests before scheduling new prefill requests

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
fastdeploy/envs.py	Adds new environment variable to configure the number of blocks to reserve per decoding request
fastdeploy/engine/sched/resource_manager_v1.py	Updates scheduling logic in two locations (WAITING and PREEMPTED request handling) to calculate and check block availability threshold that accounts for reserved blocks

[codecov-commenter]

Codecov Report

❌ Patch coverage is 29.16667% with 17 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (develop@78adf83). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
fastdeploy/engine/sched/resource_manager_v1.py	29.16%	17 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #5696   +/-   ##
==========================================
  Coverage           ?   66.97%           
==========================================
  Files              ?      347           
  Lines              ?    44474           
  Branches           ?     6834           
==========================================
  Hits               ?    29788           
  Misses             ?    12483           
  Partials           ?     2203           

Flag	Coverage Δ
GPU	66.97% <29.16%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.