v1.16.0

New: anonymous, opt-in usage telemetry — off by default, enabled via a one-time dashboard banner or Settings. One small bucketed ping a week (versions, enabled features, size buckets; never film titles, usernames, IPs, or exact numbers), a Preview-exact-JSON button showing precisely what's sent, a regenerable random instance ID, and an extra capped ping when errors start occurring so fleet-wide breakage is caught early. Full payload documented in the README. No other behaviour changes.