feat(v1.1.9): TriggerEvent::Queue + TriggerKind::Queue + types

Add the data-shape pieces for v1.1.9's queue surface; behavior lands in
later commits. Each piece compiles independently.

- shared/trigger_event.rs: TriggerEvent::Queue variant (queue_name,
  message, enqueued_at, attempt, message_id). source() returns "queue".
  Surfaced to scripts as ctx.event.queue with op = "receive".
- manager-core/trigger_repo.rs: TriggerKind::Queue + TriggerDetails::Queue +
  CreateQueueTrigger + ActiveQueueConsumer + QueueDetailRow + QueueConsumerRow.
  PostgresTriggerRepo gains create_queue_trigger (advisory-lock-then-SELECT
  enforces one-consumer-per-queue), list_active_queue_consumers (dispatcher
  scan), touch_queue_trigger_last_fired_at. hydrate_one hydrates the Queue arm.
- manager-core/outbox_repo.rs: OutboxSourceKind::Invoke for invoke_async()
  outbox rows.
- manager-core/dispatcher.rs: placeholder OutboxSourceKind::Invoke arm (logs +
  drops the row) so the workspace compiles; real arm lands in commit 10.
- manager-core/trigger_config.rs: queue_reclaim_interval_ms (30000) +
  queue_default_visibility_timeout_secs (30) env-overridable knobs.
- executor-core/engine.rs: trigger_event_to_dynamic handles Queue → builds
  ctx.event.queue map.
- manager-core/triggers_api.rs: in-memory mock TriggerRepo gains the three
  new methods (returns Default-ish values for tests).

Unit tests: TriggerEvent::Queue serde round-trip, TriggerKind::Queue wire
round-trip, advisory_lock_key stability per (app_id, queue_name) pair.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/manager-core/src/trigger_config.rs

@@ -61,6 +61,18 @@ pub struct TriggerConfig {
     /// real-world cron precision is per-minute, so a 30s tick is fine.
     /// Floored at 1s by the scheduler.
     pub cron_tick_interval_ms: u32,
+
+    /// Queue visibility-timeout reclaim task cadence, in ms (v1.1.9).
+    /// Default 30 000. The reclaim task clears claims on
+    /// `queue_messages` whose `claimed_at` exceeds the per-queue
+    /// `visibility_timeout_secs`, so a crashed consumer doesn't lose
+    /// the message.
+    pub queue_reclaim_interval_ms: u32,
+
+    /// Default per-queue visibility-timeout in seconds (v1.1.9). Used
+    /// at queue-trigger creation when the request omits an explicit
+    /// value. Default 30. The per-trigger column overrides this.
+    pub queue_default_visibility_timeout_secs: u32,
 }
 
 impl TriggerConfig {
@@ -75,6 +87,8 @@ impl TriggerConfig {
             dead_letter_retention_days: 30,
             abandoned_retention_days: 7,
             cron_tick_interval_ms: 30_000,
+            queue_reclaim_interval_ms: 30_000,
+            queue_default_visibility_timeout_secs: 30,
         }
     }
 
@@ -101,6 +115,14 @@ impl TriggerConfig {
             &mut c.cron_tick_interval_ms,
             "PICLOUD_CRON_TICK_INTERVAL_MS",
         );
+        load_u32(
+            &mut c.queue_reclaim_interval_ms,
+            "PICLOUD_QUEUE_RECLAIM_INTERVAL_MS",
+        );
+        load_u32(
+            &mut c.queue_default_visibility_timeout_secs,
+            "PICLOUD_QUEUE_DEFAULT_VISIBILITY_TIMEOUT_SECS",
+        );
         c
     }
 }