Periodically reports the node's scheduler utilization as a [:vm, :scheduler] telemetry event (measurement :utilization, percent 0-100), computed from consecutive :scheduler.sample/0 snapshots - the same scheduler_wall_time accounting observer's load charts use.

Unlike the other producers this one is stateful (utilization is the delta between two snapshots), so it runs as its own GenServer instead of a :telemetry_poller measurement.

IMPORTANT: scheduler wall time accounting adds a small permanent cost to every scheduler, so this producer is opt-in - it only starts when config :observer_web, :scheduler_utilization_poller_interval_ms is set (see Observer.Web.Telemetry). The :scheduler_wall_time system flag is enabled for the lifetime of the producer and switched back off when it terminates.