ScalableSageEncoder问题

关于ScalableSageEncoder有如下几个问题：
1. ScalableSageEncoder的使用场景能简单描述下吗？如果GNN使用了ScalableSageEncoder，是不是可以不使用conv了，因为我看ScalableSageEncoder中已经做了几层sample和aggregate操作，这样是不是和设置的conv和flow重复了？

2. store_layer为什么设置为node embedding矩阵的shape？

3. Scalable指的是什么，对于没见过的node具有可扩展性吗？

Start with the ScalableSageEncoder implementation and its surrounding usage documentation, then trace how it relates to convolution, flow, sampling, aggregation, and the store_layer shape. Done means documenting its intended use, the meaning of “Scalable,” whether it supports unseen nodes, and how it interacts with the other settings.

Written by the indexing model from the issue text.