Device-scope segmented reduce temp storage allocation issue

Device-scope segmented reduce contains the following short-circuit:

```cpp
if (num_segments <= 0)
{
return cudaSuccess;
}
```

which happens before `temp_storage_bytes` is initialized. The issue might take place:

```cpp
size_t temp_storage_bytes; // <- uninitialized
cub::DeviceSegmentedReduce::Reduce(nullptr, temp_storage_bytes);
thrust::device_vector temp_storage(temp_storage_bytes); // <- allocate random amount of memory

```

We guarantee to overwrite `temp_storage_bytes` in the docs. Therefore, we should fix this.