[BUG] dask_cudf.DataFrame.values shows chunktype of numpy ndarray

**Describe the bug**
`dask_cudf.DataFrame.values` shows `chunktype` of `numpy.ndarray` but it is actually `cupy.ndarray`.

**Steps/Code to reproduce bug**
```
import cudf
import dask_cudf
df = cudf.DataFrame({'a': list(range(20)),
'b': list(reversed(range(20))),
'c': list(range(20))})
ddf = dask_cudf.from_cudf(df, npartitions=2)
print(ddf.values)

dask.array
```
**Expected behavior**
The chunktype should be `cupy.ndarray`

**Environment overview (please complete the following information)**
- Environment location: Bare-metal
- Method of cuDF install: conda

**Environment details**
cudf `0.12.0`. cuda 10.0. python 3.7