Expose SE/DESE support to EmbeddingRocksDBWrapper for training pipeline (#4227)

Summary:

X-link: facebookresearch/FBGEMM#1304


Design doc: https://docs.google.com/document/d/149LdAEHOLP7ei4hwVVkAFXGa4N9uLs1J7efxfBZp3dY/edit?tab=t.0#heading=h.49t3yfaqmt54

Context:
We are enabling the usage of rocksDB checkpoint feature in KVTensorWrapper. This allows us to create checkpoints of the embedding tables in SSD. Later, these checkpoints are used by the checkpointing component to create a checkpoint and upload it it to the manifold 

In this diff:
1. We expose serialization and deserialization support through EmbeddingRocksDBWrapper
2. Added a function to generate logs when reading through EmbeddingRocksDB/ReadOnlyEmbeddingKVDB

Reviewed By: duduyi2013

Differential Revision: D75489895

Author: Raahul Kalyaan Jakka

fbgemm_gpu/src/ssd_split_embeddings_cache/ssd_split_table_batched_embeddings.cpp

@@ -886,7 +886,17 @@ static auto kv_tensor_wrapper =
             &KVTensorWrapper::sizes,
             std::string(
                 "Returns the shape of the original tensor. Only the narrowed part is materialized."))
-        .def_property("strides", &KVTensorWrapper::strides);
+        .def_property("strides", &KVTensorWrapper::strides)
+        .def_pickle(
+            // __getstate__
+            [](const c10::intrusive_ptr<KVTensorWrapper>& self) -> std::string {
+              return self->serialize();
+            },
+            // __setstate__
+            [](std::string data) -> c10::intrusive_ptr<KVTensorWrapper> {
+              return c10::make_intrusive<KVTensorWrapper>(data);
+            })
+        .def("logs", &KVTensorWrapper::logs, "");
 
 TORCH_LIBRARY_FRAGMENT(fbgemm, m) {
   m.def(