refactor: centralize llm backend configuration

plugins/message_summary/config.toml

@@ -4,14 +4,8 @@
 enabled = true
 
 [api]
-api_key = "app-shCA6bo5l2VDmnvhg2BtuJbk"
-api_base_url = "http://192.168.2.240/v1"
-mode = "workflow"
-endpoint = "workflows/run"
-workflow_output_key = "text"
-response_mode = "streaming"
+backend = "dify_workflow_message_summary"
 connect_timeout_seconds = 10
-request_timeout_seconds = 180
 retry_delays_seconds = [10, 20]
 
 [output]