ConvertPrecision transformation: handle Assign, ReadValue and Variable (#21266)

* ConvertPresicion transformation: handle Assign, ReadValue and Variable

* revert debug code

* use correct commit for onednn_gpu

* codestyle

src/bindings/python/tests/test_runtime/test_infer_request.py

@@ -663,9 +663,7 @@ def test_infer_queue_get_idle_handle(device):
 @pytest.mark.parametrize("data_type",
                          [np.float32,
                           np.int32,
-                          # issue after ConvertPrecision transformation, ticket: TBA
-                          # np.float16
-                          ])
+                          np.float16])
 @pytest.mark.parametrize("mode", ["set_init_memory_state", "reset_memory_state", "normal"])
 @pytest.mark.parametrize("input_shape", [[10], [10, 10], [10, 10, 10], [2, 10, 10, 10]])
 @pytest.mark.skipif(