feat: multi-language embedding with CleanTemplateText + three-branch vector strategy

- StaticEmbedder: pre-trained ConceptNet Numberbatch/fastText word embeddings,
  auto-download with TF-IDF fallback, comma-separated multi-model paths
- CleanTemplateText: regex stripping of QQ tool call templates and noise
- textForVector: per-source vector strategy (agent→Response, user→Input,
  cold_storage→both)
- Indexer.BuildContext and ExtractKeywords now clean input before vectorization
- Protect recent 10 events in Prune (regression fix: use local var not const)

internal/memory/indexer.go

@@ -90,11 +90,13 @@ func (idx *Indexer) BuildContext(userInput string) *InjectedContext {
 		return &InjectedContext{Summary: ""}
 	}
 
+	input := CleanTemplateText(userInput)
+
 	// 1. 向量搜索：从实体名向量索引中找到相关实体
-	vectorEntities := idx.vectorSearchEntities(userInput)
+	vectorEntities := idx.vectorSearchEntities(input)
 
 	// 2. 关键词搜索：已有逻辑
-	keywords := ExtractKeywords(userInput)
+	keywords := ExtractKeywords(input)
 	if len(keywords) == 0 && len(vectorEntities) == 0 {
 		keywords = []string{userInput}
 	}