[prop4n]

Increment 2 of the observability work: a persistent per-resource history, so you
can see when each resource synced, drifted, was applied or failed, and from which
commit.

Stacked on #8 (Git commit) — base this on that branch; merge #8 first.

What it does

• Store: a resource_events table in SQLite (kind, name, type, reason, commit_, at) with AppendEvent / EventsFor.
• Recording:
  • The engine records state transitions (drifted, synced, removed) by
    comparing the previous snapshot, tagged with the pass commit. A resource
    already in sync on boot is not logged (avoids a synced burst).
  • The dispatcher records applied / failed when an action finishes; the
    Action carries the commit set at plan time so events attribute correctly.
  • Recording is optional (EventSink), nil-safe; the plan command records nothing.
• API: GET /api/v1/resources/{kind}/{name}/events (newest first).
• UI: a history toggle on each resource card opens an inline timeline
  (event, reason, short commit, relative time), colour-dotted by type.
• Retention: keep-all for now (documented debt; a later increment caps it).

Tests

• Store: append + newest-first read, fields persisted.
• Engine: drift then synced emits the right transition events.
• Dispatcher: applied + failed recorded with commit/reason.
• Verified end-to-end on a live cluster: 16 real events landed in the DB
  (drifted with reason/commit/timestamp), read back through the store.

golangci-lint (v2) clean, gofmt clean, Go + vitest suites green.

Next (increment 3)

Daemon logs in the UI via an in-memory ring buffer over SSE.