Fix activity_schedule_to_start_latency reporting activity_type unexpectedly

[sachinsharma3191]

What was changed

Removed duplicate activity_schedule_to_start_latency recordings from ActivityPollTask and AsyncActivityPollTask.

Why?

Issue #2733: the metric was being emitted three times per activity task — twice at poll time with activity_type="none" (the default tag on poller-level scope), and once in ActivityWorker.handleActivity with the correct activity_type tag. After this fix, the metric is emitted exactly once per activity task with the correct activity_type, consistent with how workflow_task_schedule_to_start_latency and nexus_schedule_to_start_latency are recorded.

The root cause was that ActivityPollTask.java and AsyncActivityPollTask.java used the poller-level metricsScope, which is tagged with activity_type="none" (from MetricsTag.defaultTags). ActivityWorker.handleActivity already creates a properly-tagged scope with the real activity type.

Checklist

• Removed duplicate metric recording from ActivityPollTask.java (lines 123-127)
• Removed duplicate metric recording from AsyncActivityPollTask.java (lines 124-128)
• Verified ActivityWorker.handleActivity still records the metric with correct tags

Fixes #2733

[dplyukhin]

Thanks for the contrib. Is this PR still in progress? The checklist isn't checked off. If in progress, please convert to a draft PR. Please also update the PR description to use the template we provide. Does this address any outstanding issues? Please also take note of the contributing guidelines - particularly AI-generated PRs. Please make sure to review AI contributions thoroughly yourself before opening a PR.

[sachinsharma3191]

Thanks for the review! I've updated the PR description to use the repo's template format and checked off the checklist.

This addresses issue #2733 — removes duplicate activity_schedule_to_start_latency recordings from the poll tasks, keeping only the correctly-tagged emission in ActivityWorker.handleActivity. I've reviewed the change thoroughly.