Function 'binary_replace_slice' has no kernel matching input types (list<item: string>)

### Describe the bug, including details regarding any error messages, version, and platform.

I am unable to use `compute.binary_replace_slice` to transform a column in a batch of a dataset.

Other functions with the same signature work just fine, for example `binary_join`

To illustrate the bug

```py

import pyarrow as pa
import pyarrow.compute as pc
import pyarrow.dataset as ds

data = {
"id": ["id1", "id2"],
"video-formats": [["v11", "v12", "v12"], []],
}
table = pa.table(
{
"id": pa.array(data["id"]),
"video-formats": pa.array(data["video-formats"]),
}
)

dataset = ds.dataset(table)
for b in dataset.to_batches(batch_size=2):
for a in pc.binary_join(b["video-formats"], "#"): # this work
print(a)

dataset = ds.dataset(table)
for b in dataset.to_batches(batch_size=2):
for a in pc.binary_replace_slice(b["video-formats"], 0, 1, "#"): # this does not work
print(a)

```

### Component(s)

Python

Start with the supplied Python reproduction using pc.binary_replace_slice on the dataset batch, and compare it with the working pc.binary_join call. Trace the binary_replace_slice entry point and its kernel type matching; the issue is done when the reported list input is handled consistently with the example's expected transformation.

Written by the indexing model from the issue text.