[BUG] Grouping by a nonexistent key gives `ValueError` rather than `KeyError`

I would expect the code below to raise a `KeyError`, but it raises a `ValueError` instead:

```python
>>> import cudf
>>> df = cudf.DataFrame({'a': [1, 1, 2], 'b': [1, 2, 3]})
>>> df.groupby('cd').sum()
...
ValueError: Grouper and object must have same length
```