Update on "[ET-VK] Deserialize VkGraph in ET-VK"

Add logic to deserialize a VkGraph blob back python object. This allows us to get a implement debugging / visualization directly on the vulkan-exported program.  Still extra works need to be done: From the entire bundle, need to extract the specific vulkan delegate first.

Differential Revision: [D66443780](https://our.internmc.facebook.com/intern/diff/D66443780/)

[ghstack-poisoned]

Author: yipjustin

backends/vulkan/test/test_serialization.py

@@ -46,7 +46,7 @@ def _generate_random_const_tensors(self, num_tensors: int) -> List[torch.Tensor]
 
         return tensors
 
-    def test_serialize_vulkan_binary(self):
+    def test_serialize_vulkan_binary(self) -> None:
         vk_graph = VkGraph(
             version="0",
             chain=[],
@@ -104,7 +104,7 @@ def test_serialize_vulkan_binary(self):
             tensor_bytes = bytes(array)
             self.assertEqual(constant_data_bytes, tensor_bytes)
 
-    def test_serialize_deserialize_vkgraph(self):
+    def test_serialize_deserialize_vkgraph(self) -> None:
         in_vk_graph = VkGraph(
             version="1",
             chain=[