fix(media): 修复emby电影库混入剧集/特别篇识别/剧集合集 + 新增手动编辑元数据

emby 兼容层(hills/yamby 等客户端):
- 电影库不再混入剧集:filterLikelyEpisodicPathsFromMovieQuery 按路径
  特征(Season/S0x/Specials/电视剧/日番/国漫等)剔除误入电影库的剧集
- itemPayload 改用 mediaShouldBeEpisode:有季集号且(属剧集库或路径像剧集)
  才标记为 Episode,修正电影库里剧集被当成 Movie

特别篇识别:
- episode_parser 新增 patSpecialSeason,识别 S0/Season 0/Specials/SP/OVA/
  OAD/番外/特别篇/特典 文件夹 → 第0季;seasonFromParents 返回 found 标志,
  区分"明确第0季"与"未识别默认第1季"
- emby seasonDir 正则同步支持 Specials/SP/OVA/特别篇 等

剧集合集:
- groupSeries 抽出 isSeriesCard,卡片链接按是否剧集判定,修复剧集卡片
  跳转;每集分离的根因(季集号/分组key)随入库识别修复一并改善

手动编辑元数据(自采集视频无TMDB信息时可自定义封面/标题/简介/分类):
- 新增 MediaService.UpdateMetadata + PATCH /media/:id/metadata(需admin)
- 前端 MetadataEditDialog 接入 LibraryPage 与 MediaDetailPage

其他:scanner/scraper/nfo/local+cloud_metadata/adult 元数据识别配套改进;
.gitignore 忽略本地 .dev-cache/.dev-data 运行产物。

internal/service/cloud_path_repair.go

@@ -10,9 +10,9 @@ import (
 	"github.com/ShukeBta/MediaStationGo/internal/model"
 )
 
-// RepairCloudPathMetadata backfills external IDs from cloud paths such as
+// RepairCloudPathMetadata backfills external IDs from media paths such as
 // "Movie (2025) {tmdb-123}" so existing placeholder rows can be scraped
-// without requiring another successful cloud provider traversal.
+// without requiring another successful filesystem or cloud provider traversal.
 func (c *Container) RepairCloudPathMetadata(ctx context.Context) (int, error) {
 	if c == nil || c.Repo == nil || c.Repo.DB == nil {
 		return 0, nil
@@ -22,7 +22,6 @@ func (c *Container) RepairCloudPathMetadata(ctx context.Context) (int, error) {
 	query := c.Repo.DB.WithContext(ctx).
 		Model(&model.Media{}).
 		Select("id, title, path, year, season_num, episode_num, scrape_status, tm_db_id, bangumi_id, douban_id, thetvdb_id").
-		Where("path LIKE ?", "cloud://%").
 		Where("("+strings.Join([]string{
 			"LOWER(path) LIKE ?",
 			"LOWER(path) LIKE ?",
@@ -44,22 +43,22 @@ func (c *Container) RepairCloudPathMetadata(ctx context.Context) (int, error) {
 			updates := map[string]any{}
 			status := strings.TrimSpace(row.ScrapeStatus)
 			enrichable := status == "" || status == "pending" || status == "no_match"
-			backfilledExternalID := false
-			if meta.TMDbID > 0 && row.TMDbID <= 0 {
+			changedExternalID := false
+			if meta.TMDbID > 0 && row.TMDbID != meta.TMDbID {
 				updates["tm_db_id"] = meta.TMDbID
-				backfilledExternalID = true
+				changedExternalID = true
 			}
-			if meta.BangumiID > 0 && row.BangumiID <= 0 {
+			if meta.BangumiID > 0 && row.BangumiID != meta.BangumiID {
 				updates["bangumi_id"] = meta.BangumiID
-				backfilledExternalID = true
+				changedExternalID = true
 			}
-			if strings.TrimSpace(meta.DoubanID) != "" && strings.TrimSpace(row.DoubanID) == "" {
+			if strings.TrimSpace(meta.DoubanID) != "" && strings.TrimSpace(row.DoubanID) != strings.TrimSpace(meta.DoubanID) {
 				updates["douban_id"] = strings.TrimSpace(meta.DoubanID)
-				backfilledExternalID = true
+				changedExternalID = true
 			}
-			if strings.TrimSpace(meta.TheTVDBID) != "" && strings.TrimSpace(row.TheTVDBID) == "" {
+			if strings.TrimSpace(meta.TheTVDBID) != "" && strings.TrimSpace(row.TheTVDBID) != strings.TrimSpace(meta.TheTVDBID) {
 				updates["thetvdb_id"] = strings.TrimSpace(meta.TheTVDBID)
-				backfilledExternalID = true
+				changedExternalID = true
 			}
 			if meta.Year > 0 && row.Year <= 0 {
 				updates["year"] = meta.Year
@@ -67,7 +66,7 @@ func (c *Container) RepairCloudPathMetadata(ctx context.Context) (int, error) {
 			if enrichable && strings.TrimSpace(meta.Title) != "" && cloudPathRepairShouldReplaceTitle(row.Title, meta.Title) {
 				updates["title"] = strings.TrimSpace(meta.Title)
 			}
-			if backfilledExternalID && status == "no_match" {
+			if changedExternalID && (status == "" || status == "no_match" || status == "matched") {
 				updates["scrape_status"] = "pending"
 			}
 			if len(updates) == 0 {