feat(queue): 增加非文本模型分布式准入队列

使用 PostgreSQL 统一同步与异步非文本任务的并发准入、持久化等待和 Worker 容量分配，并将生产 API 与独立 Worker 角色拆分。

补充策略管理、共享契约、OpenAPI、Kubernetes 双节点 Worker 清单及跨节点验收脚本；未默认启用任何生产 queue_size 策略。

已在原基线完成 Go、前端、迁移、Shell、Kustomize 与长任务容量验收；合入最新主干后将重新执行发布门禁。

apps/api/internal/runner/request_context_test.go

@@ -0,0 +1,55 @@
+package runner
+
+import (
+	"context"
+	"testing"
+	"time"
+)
+
+func TestRequestExecutionContextCancelsBeforeSubmission(t *testing.T) {
+	requestCtx, cancelRequest := context.WithCancel(context.Background())
+	serverCtx, cancelServer := context.WithCancel(context.Background())
+	defer cancelServer()
+	ctx, cancel := NewRequestExecutionContext(requestCtx, serverCtx)
+	defer cancel()
+
+	cancelRequest()
+	select {
+	case <-ctx.Done():
+	case <-time.After(time.Second):
+		t.Fatal("request context did not cancel before upstream submission")
+	}
+}
+
+func TestRequestExecutionContextSurvivesDisconnectAfterSubmission(t *testing.T) {
+	requestCtx, cancelRequest := context.WithCancel(context.Background())
+	serverCtx, cancelServer := context.WithCancel(context.Background())
+	ctx, cancel := NewRequestExecutionContext(requestCtx, serverCtx)
+	defer cancel()
+
+	markUpstreamSubmissionStarted(ctx)
+	cancelRequest()
+	select {
+	case <-ctx.Done():
+		t.Fatal("client disconnect propagated after upstream submission")
+	case <-time.After(50 * time.Millisecond):
+	}
+
+	cancelServer()
+	select {
+	case <-ctx.Done():
+	case <-time.After(time.Second):
+		t.Fatal("server shutdown did not cancel detached execution")
+	}
+}
+
+func TestRequestExecutionContextPreservesRequestValues(t *testing.T) {
+	type contextKey struct{}
+	requestCtx := context.WithValue(context.Background(), contextKey{}, "trace-value")
+	ctx, cancel := NewRequestExecutionContext(requestCtx, context.Background())
+	defer cancel()
+
+	if got := ctx.Value(contextKey{}); got != "trace-value" {
+		t.Fatalf("request context value = %v, want trace-value", got)
+	}
+}