[Python] Legacy dataset can't roundtrip Int64 with nulls if partitioned

If there is partitioning and the column has nulls, Int64 columns may not round trip successfully using the legacy datasets implementation. 

Simple reproduction:

 ```python

import pyarrow as pa
import pyarrow.parquet as pq
import pyarrow.dataset as ds
import tempfile

table = pa.table({
'x': pa.array([None, 7753285016841556620]),
'y': pa.array(['a', 'b'])
})

ds_dir = tempfile.mkdtemp()
pq.write_to_dataset(table, ds_dir, partition_cols=['y'])

table_after = ds.dataset(ds_dir).to_table()
print(table['x'])
print(table_after['x'])
assert table['x'] == table_after['x']
```

```Java

[
[
null,
7753285016841556620
]
]
[
[
null
],
[
7753285016841556992
]
]
```

**Reporter**: [Will Jones](https://issues.apache.org/jira/browse/ARROW-15725) / @wjones127
#### Related issues:
- [[Python] Long-term fate of pyarrow.parquet.ParquetDataset](https://github.com/apache/arrow/issues/25775) (is related to)

_{**Note**: *This issue was originally created as [ARROW-15725](https://issues.apache.org/jira/browse/ARROW-15725). Please see the [migration documentation](https://github.com/apache/arrow/issues/14542) for further details.*}

Start with the Python reproduction using pyarrow.parquet.write_to_dataset and pyarrow.dataset, focusing on the legacy datasets implementation and partitioned Int64 data with nulls. Verify the fix by rerunning the example and confirming that table_after['x'] exactly matches table['x'], including the large integer value.

Written by the indexing model from the issue text.