Stabilize feature-factory deadlock guard test under CI load

[glen-84]

Summary

• GetOrSetSafe_Should_NotDeadlock_When_FactoryRacesLazySelectionSetCompilation intermittently hit its deadlock guard in the Main Branch Coverage workflow: under coverage instrumentation and parallel test load, thread-pool scheduling can delay one side of the test's handshake past its 5s window, and the stalled handshake was then reported as a deadlock.
• The two racing sides now run on dedicated threads (TaskCreationOptions.LongRunning), making the handshake independent of thread-pool scheduling, and the handshake/completion timeouts are raised to 30s/60s. A genuine deadlock still fails the test at the completion timeout.

Test plan

• Ran the test against the current product code: passes in under a second.
• Temporarily reverted the product fix from Fix feature factory deadlock during lazy selection-set compilation #10168 and confirmed the reworked test still detects the original deadlock, then restored the product code.

[Copilot]

Pull request overview

Stabilizes the feature-factory deadlock regression test under CI load.

Changes:

• Runs racing operations on dedicated long-running threads.
• Extends handshake and completion timeouts.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.