Improve no_poller_tasks metric

[Shivs11]

What changed?

• This metric, although debatable if it should exist or not, is incremented when there are no pollers on the root partition for 2 minutes (configurable by DC) and with tasks being added to this partition.
• A customer pointed out a race condition recently where we noticed that on pod restarts, which resulted in reloading the root partition, led to this metric being wrongly emitted. This was because it did not take into consideration pod restarts.

Why?

• Correctness

How did you test it?

• built
• run locally and tested manually
• covered by existing tests
• added new unit test(s)
• added new functional test(s)

Potential risks

• None, I think

---

Note

Prevents false positives in the no-recent-poller metric by tracking partition load time and gating emission.

• Add loadTime to taskQueuePartitionManagerImpl, set in Start()
• Emit metrics.NoRecentPollerTasksPerTaskQueueCounter only when the root partition has been loaded > noPollerThreshold and there have been no pollers in the last threshold
• Add focused unit tests using metricstest.Capture covering newly loaded vs. older partitions and with/without recent pollers
• Minor wiring: import metricstest, helper to setup a manager with capturing metrics

^{Written by Cursor Bugbot for commit b6380e5. This will update automatically on new commits. Configure here.}