[task]DeviceAddress数据结构优化

**持续优化DeviceAddress数据结构** **目标：** shape type format信息上移到kernel_tensor和tensor中 **Sync接口整改** 1. device_address.h 提供DeviceAddressInfo结构： struct DeviceAddressInfo { bool is_enable{false}; Format format_{Format::DEFAULT_FORMAT}; TypeId dtype_id_{kTypeUnknown}; ShapeVector shape_vector_{}; }; 2. 底层copy接口增加DeviceAddressInfo类型参数： MS_CORE_API bool SyncCopy(const DeviceAddressPtr &dst_device_address, const DeviceAddressPtr &src_device_address, size_t stream_id, **const DeviceAddressInfo &src_info = {}, const DeviceAddressInfo &dst_info = {}** ); MS_CORE_API bool AsyncCopy(const DeviceAddressPtr &dst_device_address, const DeviceAddressPtr &src_device_address, size_t stream_id, bool keep_src = true, **const DeviceAddressInfo &src_info = {}, const DeviceAddressInfo &dst_info = {}** ); 3. 提供一套高阶拷贝接口： MS_CORE_API bool SyncCopy(const tensor::TensorPtr &dst, const tensor::TensorPtr &src, size_t stream_id); MS_CORE_API bool AsyncCopy(const tensor::TensorPtr &dst, const tensor::TensorPtr &src, size_t stream_id, bool keep_src = true); RUNTIME_HARDWARE_EXPORT bool SyncCopy(kernel::KernelTensor *const dst_kernel_tensor, kernel::KernelTensor *const src_kernel_tensor, size_t stream_id); RUNTIME_HARDWARE_EXPORT bool SyncCopy(const tensor::TensorPtr &dst_tensor, kernel::KernelTensor *const src_kernel_tensor, size_t stream_id); RUNTIME_HARDWARE_EXPORT bool SyncCopy(kernel::KernelTensor *const dst_kernel_tensor, tensor::Tensor *const src_tensor, size_t stream_id); RUNTIME_HARDWARE_EXPORT bool AsyncCopy(kernel::KernelTensor *const dst_kernel_tensor, kernel::KernelTensor *const src_kernel_tensor, size_t stream_id, bool keep_src = true); RUNTIME_HARDWARE_EXPORT bool AsyncCopy(kernel::Kerne…