P2P shuffling failed during transfer when data includes mixed type

When there are object columns with mixed data types, the arrow backend cannot handle this.

The internal error that is raised in this example is

```python-traceback
ArrowInvalid: ('cannot mix list and non-list, non-null values', 'Conversion failed for column mixed_stuff with type object')
```

while the user receives a generic shuffle failed exception

```python-traceback
RuntimeError: P2P shuffling [id] failed during transfer phase
```

Reproducing code example

```python
import pandas as pd
import dask.dataframe as dd
import numpy as np
from distributed import Client
with Client() as client:

df = pd.DataFrame({
"mixed_stuff": [{"foo": "bar"}, np.array((3,))] * 2,
"int": [1, 2] * 2,
})
ddf = dd.from_pandas(df, npartitions=2)
ddf.shuffle(on="int").compute()
```