Supports over window on both limited and timed window in SparkProcessor

over_window_utils.py:

```python
def _get_spark_window_spec(
over_window_descriptor: "OverWindowDescriptor",
) -> WindowSpec:
if (
over_window_descriptor.limit is not None
and over_window_descriptor.window_size is not None
):
# TODO Supports aggregations on both limited and timed window
raise FeathubException(
"You cannot set window_size and limit of over window at the same time."
)
```

test_spark_processor.py

```python
# TODO: Add back following test cases after SparkProcessor supports aggregations
# on both limited and timed window
def test_over_window_on_join_field(self):
pass
```

Read over_window_utils.py, especially _get_spark_window_spec, then inspect test_spark_processor.py and its TODO test_over_window_on_join_field case. Run the relevant SparkProcessor tests; done means aggregations can use both limited and timed windows, with coverage for the join-field case.

Written by the indexing model from the issue text.