Throttle ZFS collector when no UI is connected

[gnacho]

Context

The periodic ZFS collector is the main source of sudo/journald log volume on nodes that run EasyZFS 24/7. We already reduced log volume in #124 by lowering the collector log level, but the collector still runs every EASYZFS_ZPOOL_INTERVAL (default 60 s) regardless of whether a user is looking at the UI.

The web UI connects to /api/events and keeps an SSE stream open while it is in use. The Hub already tracks active subscribers, so we can use that as a cheap presence signal.

Goal

• Run a cheap health/alert heartbeat every ~60 s even when the UI is closed, so background alerts (DEGRADED, scrub, capacity) still fire promptly.
• Run a full data collection only when the UI is connected or, at worst, every ~5 minutes, so datasets/snapshots/history/series are still refreshed but the node is quiet most of the time.
• Keep this configurable through environment variables with sensible defaults.

Proposed design

1. Add SubscriberCount() to internal/hub/hub.go.
2. Pass a presence callback to NewZpoolCollector (or read it from the existing Hub).
3. Split collectOnce into:
   • light collect (zpool list, zpool status --json per pool): evaluate alerts, publish state changes.
   • full collect (datasets, snapshots, history, series): reuse existing logic.
4. Dynamic schedule:
   • UI connected: full collect every EASYZFS_ZPOOL_INTERVAL (default 10 s).
   • UI disconnected: light collect every EASYZFS_ZPOOL_ALERT_INTERVAL (default 60 s), full collect every EASYZFS_ZPOOL_IDLE_INTERVAL (default 300 s).
5. Add tests for the new interval logic.

Acceptance criteria

• zpool status / zpool list commands are dramatically reduced when no UI client is connected.
• Alerts still evaluate at least every minute without a UI.
• Historical series still persist at least every 10 minutes.
• Existing EASYZFS_ZPOOL_INTERVAL keeps working as the UI open interval.
• New env vars documented in README.md / README.es.md.
• Release and deploy v2.9.21.

Related

• Reduce sudo/journald log volume from periodic collectors #124 / fix: reduce sudo/journald log volume from periodic collectors (#124) #125 — reduced sudo log volume from periodic collectors.