fix(kling): 补齐全模态基础视频能力

让 omni_video 模型默认同时注册 video_generate 与 image_to_video，兼容通用视频接口在不显式传 modelType 时的能力推断。

新增迁移同步可灵 O1 与 3.0 Omni 的基础目录、平台映射、能力详情和默认快照；集成测试覆盖两个模型的无 modelType 文生视频与图生视频。

apps/api/internal/store/candidates_test.go

@@ -9,6 +9,19 @@ func TestNormalizeModelMatchKeyRemovesWhitespace(t *testing.T) {
 	}
 }
 
+func TestNormalizeModelTypeListExpandsOmniVideoBaseCapabilities(t *testing.T) {
+	got := normalizeModelTypeList([]string{"omni_video"})
+	want := StringList{"video_generate", "image_to_video", "omni_video"}
+	if len(got) != len(want) {
+		t.Fatalf("omni_video should include text-to-video and image-to-video capabilities: got=%v want=%v", got, want)
+	}
+	for index := range want {
+		if got[index] != want[index] {
+			t.Fatalf("omni_video capability mismatch at %d: got=%v want=%v", index, got, want)
+		}
+	}
+}
+
 func TestTaskBillingModelIdentityKeepsRequestedModelPrimary(t *testing.T) {
 	identity := taskBillingModelIdentity(GatewayTask{
 		Model:          "doubao-5.0 图像编辑",