feat(alerts): alert when a known pool is not imported

[gnacho]

Closes #136.

What

EvaluatePools only evaluated pools returned by the collector. When a pool fails to auto-import (e.g. after an unclean shutdown), it vanishes from zpool list and no alert was raised, even though every dataset and container on it stays down.

How

• New known_pools table (migration v23): the alerter persists every pool it sees, with first/last seen timestamps.
• After each collection (light and full), a known pool absent for more than EASYZFS_POOL_MISSING_AFTER seconds (default 300) raises a critical pool_missing alert. Fresh installs (empty table) never alert, so "never seen" is distinguished from "seen before and now missing".
• New kind pool_missing in the push/email/channel i18n catalog (ES/EN) and in the notification preference types.
• New GET /api/pools/missing endpoint; the dashboard shows a persistent critical banner listing the missing pools with a shortcut to Pools.

Detection is strictly read-only: the app never attempts to import.

Test plan

• 5 new tests in internal/alerts/pool_missing_test.go (fresh install, missing pool alerts, grace window, re-import without re-alert, API contract). Full suite green.
• Validated live on two hosts: seeded a stale pool entry, alert + banner appeared; removed it, state back to normal.