Decouple metric fetch errors from trial status in Orchestrator (#5287)

[saitcakmak]

Summary:

Keep trial execution status authoritative to runner polling when metric fetching fails. Orchestrator now leaves completed trials completed, uses MetricAvailability to count incomplete optimization-config data in the existing failure-rate policy, and handles expanded multi-metric result keys without assuming they are registered experiment metrics.

This mirrors the behavior core Ax already has in the Orchestrator.

This implements the core design from the abandoned RFC D98741656 and fixes T287759832.

Reviewed By: esantorella

Differential Revision: D119207750

[meta-codesync]

@saitcakmak has exported this pull request. If you are a Meta employee, you can view the originating Diff in D119207750.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.59%. Comparing base (1cb1f8a) to head (8a51f43).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5287      +/-   ##
==========================================
+ Coverage   96.57%   96.59%   +0.02%     
==========================================
  Files         621      621              
  Lines       70947    70958      +11     
==========================================
+ Hits        68518    68544      +26     
+ Misses       2429     2414      -15     

☔ View full report in Codecov by Harness.
📢 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.