feat(saas): 接通 embedding 模型管理全栈

数据库 migration 已有 is_embedding/model_type 列但全栈未使用。
打通 4 层: ModelRow → ModelInfo/CRUD → CachedModel → Admin 前端。
relay/models 端点也返回 is_embedding 字段，前端可按类型过滤。

crates/zclaw-saas/src/relay/handlers.rs

@@ -373,9 +373,10 @@ pub async fn list_available_models(
     _ctx: Extension<AuthContext>,
 ) -> SaasResult<Json<Vec<serde_json::Value>>> {
     // 单次 JOIN 查询替代 2 次全量加载
-    let rows: Vec<(String, String, String, i64, i64, bool, bool)> = sqlx::query_as(
+    let rows: Vec<(String, String, String, i64, i64, bool, bool, bool, String)> = sqlx::query_as(
         "SELECT m.model_id, m.provider_id, m.alias, m.context_window,
-                m.max_output_tokens, m.supports_streaming, m.supports_vision
+                m.max_output_tokens, m.supports_streaming, m.supports_vision,
+                m.is_embedding, m.model_type
          FROM models m
          INNER JOIN providers p ON m.provider_id = p.id
          WHERE m.enabled = true AND p.enabled = true
@@ -385,7 +386,7 @@ pub async fn list_available_models(
     .await?;
 
     let mut available: Vec<serde_json::Value> = rows.into_iter()
-        .map(|(model_id, provider_id, alias, context_window, max_output_tokens, supports_streaming, supports_vision)| {
+        .map(|(model_id, provider_id, alias, context_window, max_output_tokens, supports_streaming, supports_vision, is_embedding, model_type)| {
             serde_json::json!({
                 "id": model_id,
                 "provider_id": provider_id,
@@ -394,6 +395,8 @@ pub async fn list_available_models(
                 "max_output_tokens": max_output_tokens,
                 "supports_streaming": supports_streaming,
                 "supports_vision": supports_vision,
+                "is_embedding": is_embedding,
+                "model_type": model_type,
             })
         })
         .collect();