[Scheduler] Add the fused task name to auto-scheduled kernels (#418)

This PR makes the generated kernel name contained more information.
Previously, the name might look like `hidet_compute_c`, now it might
look like `hidet_fused_softmax_rearrange_c`.

python/hidet/ir/schedulers/cuda/scheduler.py

@@ -35,7 +35,13 @@ def schedule_grid_compute(self, node: GridCompute, tensor_map: Dict[TensorNode,
         grid_dim: Expr = (prod(node.shape) + block_dim - 1) // block_dim
 
         if self.task is not None:
-            name = f'{self.task.name}_compute_{node.name}'
+            from hidet.graph.ops.fusion.fused_operator import FusedTask
+
+            if isinstance(self.task, FusedTask):
+                fused_name = self.task.attrs['fused_ops'].replace(' ', '_')
+                name = f'fused_{fused_name}_{node.name}'
+            else:
+                name = f'{self.task.name}_{node.name}'
         else:
             name = f'compute_{node.name}'