feat(server): hot-reload config store and reloadable infra clients

- Replace single Load() with a watching Store that re-reads config files
  (and config.local.yaml) and fans out a ReloadEvent with a per-field diff
  so consumers can decide whether the change is hot-applicable or requires
  a process restart.
- Wrap llm, retrieval, vector store, and object storage clients in
  Reloadable* shells so the bootstrap can swap their underlying impls when
  config changes without re-instantiating handlers.
- Make jwt.Manager and ops TokenIssuer mutable under a lock so secrets and
  TTLs can be rotated live; thread default plan code through a setter on
  the ops AdminUserService.
- Wire ConfigStore through bootstrap and every cmd/main.go, scheduler /
  worker / tenant-api / ops-api start the watcher; services and handlers
  take a config.Provider so they always read current values for things
  like generation.stream_enabled, scheduler dispatch, retrieval, etc.
- Switch shared/config decoding off viper to a Kratos-derived runtime
  package so env placeholders (\${VAR:default}) resolve consistently and
  the same source machinery powers both the loader and the watcher.

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

server/internal/tenant/transport/article_handler.go

@@ -24,7 +24,7 @@ type ArticleHandler struct {
 	promptGenerateSvc *app.PromptRuleGenerationService
 	imitationSvc      *app.ArticleImitationService
 	streams           *stream.GenerationHub
-	streamEnabled     bool
+	app               *bootstrap.App
 	assets            *AssetHandler
 }
 
@@ -36,40 +36,40 @@ func NewArticleHandler(a *bootstrap.App) *ArticleHandler {
 		a.ObjectStorage,
 		a.LLM,
 		a.Logger,
-		a.Config.Retrieval,
-		a.Config.LLM,
+		a.Config().Retrieval,
+		a.Config().LLM,
 		0,
 		0,
-	)
+	).WithConfigProvider(a.ConfigStore)
 
 	return &ArticleHandler{
 		svc: app.NewArticleService(a.DB, a.AuditLogs, a.ObjectStorage).WithCache(a.Cache),
 		selectionOptimize: app.NewArticleSelectionOptimizeService(
 			a.DB,
 			a.LLM,
-			a.Config.LLM.MaxOutputTokens,
-		).WithCache(a.Cache),
+			a.Config().LLM.MaxOutputTokens,
+		).WithCache(a.Cache).WithConfigProvider(a.ConfigStore),
 		promptGenerateSvc: app.NewPromptRuleGenerationService(
 			a.DB,
 			a.LLM,
 			a.RabbitMQ,
 			knowledgeSvc,
 			a.GenerationStreams,
-			a.Config.Generation,
-			a.Config.LLM.MaxOutputTokens,
-		).WithCache(a.Cache).WithLogger(a.Logger),
+			a.Config().Generation,
+			a.Config().LLM.MaxOutputTokens,
+		).WithCache(a.Cache).WithLogger(a.Logger).WithConfigProvider(a.ConfigStore),
 		imitationSvc: app.NewArticleImitationService(
 			a.DB,
 			a.LLM,
 			a.RabbitMQ,
 			knowledgeSvc,
 			a.GenerationStreams,
-			a.Config.Generation,
-			a.Config.LLM.MaxOutputTokens,
-		).WithCache(a.Cache),
-		streams:       a.GenerationStreams,
-		streamEnabled: a.Config.Generation.StreamEnabled,
-		assets:        NewAssetHandler(a),
+			a.Config().Generation,
+			a.Config().LLM.MaxOutputTokens,
+		).WithCache(a.Cache).WithConfigProvider(a.ConfigStore),
+		streams: a.GenerationStreams,
+		app:     a,
+		assets:  NewAssetHandler(a),
 	}
 }
 
@@ -396,7 +396,8 @@ func (h *ArticleHandler) OptimizeSelectionStream(c *gin.Context) {
 }
 
 func (h *ArticleHandler) Stream(c *gin.Context) {
-	if !h.streamEnabled {
+	cfg := h.app.Config()
+	if cfg == nil || !cfg.Generation.StreamEnabled {
 		response.Error(c, response.ErrNotFound(40412, "generation_stream_disabled", "generation stream is disabled"))
 		return
 	}