fix unit testing for conv layers and add correct mat multiplication i…

…n conv layer

topomodelx/base/conv.py

@@ -118,7 +118,7 @@ def forward(self, x_source, neighborhood, x_target=None):
             )
 
         x_message = torch.mm(x_source, self.weight)
-        x_message_on_target = torch.mm(neighborhood, x_message)
+        x_message_on_target = torch.sparse.mm(neighborhood, x_message)
 
         if self.aggr_norm:
             neighborhood_size = torch.sum(neighborhood.to_dense(), dim=1)