DAOS-16355 pydaos: fix dir obj cache invalid handle error (#15997)

Directory objects lifecycle is defined by cache, releasing object outside it 
will result in the error: Invalid object handler.

Signed-off-by: Denis Barakhtanov <dbarahtanov@enakta.com>

Author: 0xE0F

src/client/pydaos/torch/torch_shim.c

@@ -556,10 +556,6 @@ __shim_handle__torch_list_with_anchor(PyObject *self, PyObject *args)
 	D_FREE(dentries);
 	D_FREE(stats);
 
-	if (obj) {
-		rc = dfs_release(obj);
-	}
-
 	return PyLong_FromLong(rc);
 }
 

src/tests/ftest/pytorch/dataset.py

@@ -41,14 +41,18 @@ def test_map_style_dataset(self):
         subdirs = self.params.get("subdirs", "/run/map_style_dataset/*")
         files_per_node = self.params.get("files_per_node", "/run/map_style_dataset/*")
         file_min_size = self.params.get("file_min_size", "/run/map_style_dataset/*", 4096)
-        file_max_size = self.params.get("file_max_size", "/run/map_style_dataset/*", 128 * 1024)
+        file_max_size = self.params.get("file_max_size", "/run/map_style_dataset/*", 4096)
+        readdir_batch_size = self.params.get("readdir_batch_size", "/run/map_style_dataset/*", 5)
 
         self._create_test_files(root_dir, height, subdirs, files_per_node,
                                 file_min_size, file_max_size)
 
         expected = self._get_test_files_hashmap(root_dir, self.hostlist_clients)
 
-        dataset = Dataset(pool.identifier, container.identifier)
+        dataset = Dataset(pool=pool.identifier,
+                          cont=container.identifier,
+                          readdir_batch_size=readdir_batch_size,
+                          )
 
         actual = {}
         for _, content in enumerate(dataset):

src/tests/ftest/pytorch/dataset.yaml

@@ -1,5 +1,5 @@
 hosts:
-  test_servers: 1
+  test_servers: 2
   test_clients: 1
 server_config:
   name: daos_server
@@ -18,17 +18,18 @@ pool:
 container:
   type: POSIX
   control_method: daos
+  dir_oclass: SX
 
-timeout: 180
+timeout: 270
 
 map_style_dataset:
-  tree_height: 4
-  subdirs: 3
-  files_per_node: 5
+  tree_height: 2
+  subdirs: 2
+  files_per_node: 64
 
 iterable_dataset:
-  tree_height: 3
-  subdirs: 3
+  tree_height: 4
+  subdirs: 5
   files_per_node: 6
 
 map_dataset_with_dataloader: