Add CheckInputDataframe to DropSimilar to avoid issues with column names that aren't unique strings

The DropSimilar transformer fails if the dataframe to transform includes non-unique strings, or non-string column names.

We may want to include a checkinputdataframe to ensure column names are unique strings to avoid issues.

```python
import numpy as np
import skrub
import pandas as pd

df = pd.DataFrame(np.eye(3))
skrub.DropSimilar().fit_transform(df)
```