Summary

• remove the unused last_idle_ms bookkeeping from the event-driven capture loop
• stop calling ActivityFeed::idle_ms() during each timer-only poll when the result is never read
• refresh a stale comment about repeated idle capture attempts

Why

The event-driven capture loop wakes periodically even when idle so it can service timer-driven captures and visual checks. The old poll_activity path still performed an activity-feed wall-clock read on every wake, but the stored value was dead state after typing/scroll pause triggers moved to the UI recorder. Dropping that read shaves a small amount of steady idle work without changing capture timing or event-triggered behavior.

Validation

• cargo fmt --package screenpipe-engine -- --check
• cargo test -p screenpipe-engine event_driven_capture --lib
• cargo check -p screenpipe-engine