Optimize graphrag again (#6513)

### What problem does this PR solve?

Removed set_entity and set_relation to avoid accessing doc engine during
graph computation.
Introduced GraphChange to avoid writing unchanged chunks.

### Type of change

- [x] Performance Improvement

api/ragflow_server.py

@@ -47,6 +47,8 @@ from rag.utils.redis_conn import RedisDistributedLock
 
 stop_event = threading.Event()
 
+RAGFLOW_DEBUGPY_LISTEN = int(os.environ.get('RAGFLOW_DEBUGPY_LISTEN', "0"))
+
 def update_progress():
     lock_value = str(uuid.uuid4())
     redis_lock = RedisDistributedLock("update_progress", lock_value=lock_value, timeout=60)
@@ -85,6 +87,11 @@ if __name__ == '__main__':
     settings.init_settings()
     print_rag_settings()
 
+    if RAGFLOW_DEBUGPY_LISTEN > 0:
+        logging.info(f"debugpy listen on {RAGFLOW_DEBUGPY_LISTEN}")
+        import debugpy
+        debugpy.listen(("0.0.0.0", RAGFLOW_DEBUGPY_LISTEN))
+
     # init db
     init_web_db()
     init_web_data()