feat(memory): hybrid vector search via embedder

- repository: optional pgvector embedding column (migration with
  CREATE EXTENSION guarded by main), SearchByVector cosine search,
  vector-enabled insert path
- service: SetEmbedder — memories are embedded on Add; Search merges
  vector hits with n-gram keyword matches (vector first, deduped)
- embedder was already implemented (OpenAI-compatible API + n-gram FNV
  fallback) but never instantiated; main now wires it from
  EMBEDDING_API_URL/EMBEDDING_MODEL config

backend/services/memory/model.go

@@ -19,6 +19,10 @@ type Memory struct {
 	LastAcc   time.Time `json:"last_accessed_at"`
 	CreatedAt time.Time `json:"created_at"`
 	UpdatedAt time.Time `json:"updated_at"`
+
+	// Embedding is the vector representation used for semantic search.
+	// Not serialized or scanned by default queries; pgvector-only.
+	Embedding []float64 `json:"-"`
 }
 
 // SearchQuery represents a memory search request.