Datetime index errors in DummyEncoder and OrdinalEncoder

I've noticed that inverse_transform() is trying to iterate through the dataset by index and thus raises:

`ValueError: Cannot add integral value to Timestamp without freq.`

A small example below for the DummyEncoder, although the same thing happens with the OrdinalEncoder:

The dask dataframe:


Then I pass the intended feature column to the Categorizer:
```
cat = Categorizer(columns=['float_feature'])
df_categorized = cat.fit_transform(df)
```
At this point I setup the transformer and transform the dataframe's selected columns:
```
dummy_encoder = DummyEncoder(columns=['float_feature'])
dummy_encoder2 = dummy_encoder.fit(df_categorized)
dummy_encoded = dummy_encoder2.transform(df_categorized)
```

Finally the inverse_transform:


Why is the index relevant to these encoders' inverse transforms? Is this the intended behaviour?

Update: For clarity and minor orthography issues.