make xpu ops compatible with oneapi 2025.0 (#6760)

Compatibility update for xpu ops

This PR introduces changes that will make xpu ops compatible with the
OneAPI 2025.0 toolkit. This is an important update that will allow us to
develop and ship our most demanding models on this innovative hardware.

---------

Signed-off-by: baodii <di.bao@intel.com>
Co-authored-by: Logan Adams <114770087+loadams@users.noreply.github.com>
Co-authored-by: Logan Adams <loadams@microsoft.com>

csrc/xpu/adam/multi_tensor_apply.dp.hpp

@@ -109,6 +109,12 @@ private:
     std::tuple<ArgTypes...> args;
 };
 
+// to make sure multi_tensor_apply_kernel can be used in sycl::buffer
+namespace sycl {
+template <typename T, typename U, typename... ArgTypes>
+struct is_device_copyable<multi_tensor_apply_kernel<T, U, ArgTypes...>> : std::true_type {};
+}  // namespace sycl
+
 template <int depth, typename T, typename... ArgTypes>
 void multi_tensor_apply(int block_size,
                         int chunk_size,