feat(tasks): 任务并发上限(可选,超限排队自动补位)

任务页新建任务左侧加「并发设置」齿轮:开关(默认关) + 并发数量(开启默认5)。
开启后新建任务若「运行中」数已达上限,进入「排队中」不启动;有空位时后台
每 5s 自动把最早排队的任务补位启动。

- db: tasks 加持久化 queued 列(schema + ALTER 补旧库);Task.Queued/SetQueued。
- manager: ConcurrencyLimit/SetConcurrency(settings KV);SetTaskQueued(落库+同步内存)。
- 准入与促进器(goals.go):launchTask 统一准入,满则挂起排队;抽出 startTaskEngine
  复用;occupiesSlot/runningTaskCount/queueIfAtCapacity;reconcileConcurrency 每
  scheduler tick 补位(关闭则放行全部排队),concMu 串行化防超发;启动重载/setActive/
  reviveTask 三处'排队则不启动'门。
- settings: /api/settings 增 task_concurrency_enabled/limit;PUT 后立即协调一次。
- web: 齿轮 Dialog(Switch+数量);新增 queued/「排队中」状态(类型/状态色/筛选)。
- 超限只排队新任务,不打断在跑的;排队任务不消耗 LLM(目标拆解推迟到出队)。

go build/vet、tsc 通过;DB queued 往返测试通过。

db/tasks.go

@@ -13,6 +13,7 @@ type Task struct {
 	ExplorationID int64      `json:"exploration_id"`
 	Status        string     `json:"status"`
 	Paused        bool       `json:"paused"`
+	Queued        bool       `json:"queued"` // 因并发上限被挂起、等待空位自动启动;true=尚未开跑
 	LLMProfileID  *int64     `json:"llm_profile_id,omitempty"`
 	ParentRef     string     `json:"parent_ref,omitempty"` // 父任务 id(编排 spawn 记录;空=顶层)
 	CreatedAt     time.Time  `json:"created_at"`
@@ -84,11 +85,11 @@ RETURNING id, status, paused, created_at`, description, goal, expID, llmProfileI
 	return t, tx.Commit()
 }
 
-const taskCols = `id, description, goal, exploration_id, status, paused, llm_profile_id, COALESCE(parent_ref,''), created_at, completed_at, COALESCE(timeout_seconds,0), COALESCE(plan_heartbeat_seconds,300), first_run_at, deadline_at`
+const taskCols = `id, description, goal, exploration_id, status, paused, queued, llm_profile_id, COALESCE(parent_ref,''), created_at, completed_at, COALESCE(timeout_seconds,0), COALESCE(plan_heartbeat_seconds,300), first_run_at, deadline_at`
 
 func scanTask(sc interface{ Scan(...any) error }) (*Task, error) {
 	var t Task
-	if err := sc.Scan(&t.ID, &t.Description, &t.Goal, &t.ExplorationID, &t.Status, &t.Paused, &t.LLMProfileID, &t.ParentRef, &t.CreatedAt, &t.CompletedAt, &t.TimeoutSeconds, &t.PlanHeartbeatSeconds, &t.FirstRunAt, &t.DeadlineAt); err != nil {
+	if err := sc.Scan(&t.ID, &t.Description, &t.Goal, &t.ExplorationID, &t.Status, &t.Paused, &t.Queued, &t.LLMProfileID, &t.ParentRef, &t.CreatedAt, &t.CompletedAt, &t.TimeoutSeconds, &t.PlanHeartbeatSeconds, &t.FirstRunAt, &t.DeadlineAt); err != nil {
 		return nil, err
 	}
 	return &t, nil
@@ -136,6 +137,13 @@ func (d *DB) SetPaused(id int64, paused bool) error {
 	return err
 }
 
+// SetQueued persists a task's queued flag (并发上限挂起态). A queued task is held
+// (engine not started) until the concurrency reconciler promotes it into a free slot.
+func (d *DB) SetQueued(id int64, queued bool) error {
+	_, err := d.Exec(`UPDATE tasks SET queued=$1 WHERE id=$2`, queued, id)
+	return err
+}
+
 // SetStatus updates a task's lifecycle status. Entering a terminal state
 // (done/failed/timeout) stamps completed_at once (COALESCE keeps the first stamp
 // stable); moving back to a non-terminal state clears it, so a re-run has no stale