[Cherry-Pick] [BugFix] fix speculative gauge metrics in multi api server (#7082)

[liyonghua0910]

Motivation

This PR fixes duplicated/incorrect speculative gauge metrics export under MultiAPIServer in multi-DP scenarios.

There are two issues:

1. speculative gauge metrics were not fully filtered and re-registered during multiprocess metrics export, which could expose duplicated values or stale values from the multiprocess collector.
2. MultiAPIServer did not isolate PROMETHEUS_MULTIPROC_DIR across DP instances, which could cause metric file collisions between different DP processes.

Modifications

1. Fix the speculative metrics export logic in multiprocess mode:

   • correctly filter speculative gauge metrics from the multiprocess collector output
   • re-register speculative gauge metrics from in-memory main_process_metrics
   • cover both scalar speculative gauges and single-head speculative gauge metrics

2. Isolate PROMETHEUS_MULTIPROC_DIR for each DP instance under MultiAPIServer:

   • ensure each DP process uses an independent multiprocess metrics directory
   • avoid cross-DP metric pollution and duplicated gauge samples

3. Add unit tests for multiprocess gauge export:

   • regular gauge returns a single value without pid
   • speculative gauge returns a single value without pid
   • speculative single-head gauge returns a single value without pid

Usage or Command

Run unit tests:

pytest -q tests/metrics/test_metrics.py

Accuracy Tests

N/A. This PR does not change model computation or model outputs. It only fixes metrics export behavior.

Checklist

• Add at least a tag in the PR title.
• 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!

[codecov-commenter]

Codecov Report

❌ Patch coverage is 57.57576% with 14 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (release/2.5@054add2). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
fastdeploy/metrics/metrics.py	50.00%	9 Missing and 4 partials ⚠️
fastdeploy/entrypoints/openai/multi_api_server.py	85.71%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@              Coverage Diff               @@
##             release/2.5    #7100   +/-   ##
==============================================
  Coverage               ?   69.07%           
==============================================
  Files                  ?      390           
  Lines                  ?    54109           
  Branches               ?     8525           
==============================================
  Hits                   ?    37377           
  Misses                 ?    14022           
  Partials               ?     2710           

Flag	Coverage Δ
GPU	69.07% <57.57%> (?)

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.