feat(server/diagnostics): device debug-reporting ingest + admin timeline + retention (M9)

New diagnostic_events table + per-account users.debug_mode_enabled flag.
When an account's flag is on, its client(s) POST a batch timeseries of
connectivity / UPnP-sync / power / lifecycle events to /api/diagnostics
(no-op 204 when off, kind whitelist mirrors the CHECK constraint).

Admin surface: GET /api/admin/diagnostics (optional account/device/kind/
time-window filters, RFC3339-or-epoch-ms, export-sized paging) + a
/diagnostics/devices overview + PUT /api/admin/users/{id}/debug-mode to
flip an account remotely while a bug is live. debug_mode_enabled is now
exposed on /api/me (client gate) and the admin user views.

Retention: a 30-day gc-worker sweep (GcPruneDiagnostics), keyed on the
server clock so a skewed device clock can't keep rows alive.

Refs Scribe M9 (#119), tasks #1172 #1173.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01K55iTxn95BtshocgdE1shW

internal/db/migrations/0036_diagnostics.down.sql

@@ -0,0 +1,2 @@
+ALTER TABLE users DROP COLUMN IF EXISTS debug_mode_enabled;
+DROP TABLE IF EXISTS diagnostic_events;

internal/db/migrations/0036_diagnostics.up.sql

@@ -0,0 +1,54 @@
+-- Device diagnostics / debug-reporting subsystem (M9).
+--
+-- When an account has users.debug_mode_enabled = true, that account's
+-- client(s) stream a timeseries of device state here: connectivity +
+-- server-health transitions (roaming / poor-data recovery), UPnP/Sonos
+-- sync events (locked-phone desync), and power/lifecycle state (Doze /
+-- battery-optimization, the prime suspect for the desync). The admin
+-- /admin/diagnostics view filters by account + device + time window and
+-- exports the slice as JSON for offline analysis.
+--
+-- `kind` is a COARSE category gated by a CHECK whitelist; the finer
+-- event discriminator + all event-specific fields live in the `payload`
+-- jsonb. This keeps the enum small and stable so new event variants
+-- don't require a migration (per the enum-CHECK-whitelist rule, only a
+-- new *category* would). If you add a category here, mirror it in the
+-- handler whitelist in internal/api/diagnostics.go.
+--
+-- Two timestamps on purpose: occurred_at is the client device clock
+-- (when the event happened on the phone — may be skewed/wrong, e.g. in
+-- a dead zone), received_at is the server clock at ingest. Retention
+-- pruning is keyed on received_at so a bad client clock can't keep rows
+-- alive forever.
+
+CREATE TABLE diagnostic_events (
+    id          uuid PRIMARY KEY DEFAULT gen_random_uuid(),
+    user_id     uuid NOT NULL REFERENCES users(id) ON DELETE CASCADE,
+    client_id   text NOT NULL,
+    app_version text,
+    os_version  text,
+    kind        text NOT NULL,
+    payload     jsonb NOT NULL DEFAULT '{}'::jsonb,
+    occurred_at timestamptz NOT NULL,
+    received_at timestamptz NOT NULL DEFAULT now(),
+    CONSTRAINT diagnostic_events_kind_check
+        CHECK (kind IN (
+            'connectivity', 'upnp_sync', 'power', 'lifecycle',
+            'heartbeat', 'http'
+        ))
+);
+
+-- Admin reads filter by account then device, ordered newest-first.
+CREATE INDEX idx_diagnostic_events_user
+    ON diagnostic_events (user_id, occurred_at DESC);
+CREATE INDEX idx_diagnostic_events_client
+    ON diagnostic_events (client_id, occurred_at DESC);
+-- Retention pruner deletes by server-clock age.
+CREATE INDEX idx_diagnostic_events_received
+    ON diagnostic_events (received_at);
+
+-- Per-account opt-in for diagnostics reporting. Admin-set primarily
+-- (flip remotely while a bug is live); the client also exposes a local
+-- OFF switch. Default false: no account reports until explicitly enabled.
+ALTER TABLE users
+    ADD COLUMN IF NOT EXISTS debug_mode_enabled boolean NOT NULL DEFAULT false;