SparkConnect's notion of the Spark version isn't tied to the actual image

### What happened?

The SparkConnect CR construction utility allows the spark version to be configurable.

https://github.com/kubeflow/sdk/blob/3febcd7377d0cdb322bbb420207c3c26cbf7afd7/kubeflow/spark/backends/kubernetes/utils.py#L512

At the same time it's only caller does not leverage that provision.

https://github.com/kubeflow/sdk/blob/3febcd7377d0cdb322bbb420207c3c26cbf7afd7/kubeflow/spark/backends/kubernetes/backend.py#L151-L161

This provision is therefore a no-op and the spark version is always defaulted to `DEFAULT_SPARK_VERSION`.

https://github.com/kubeflow/sdk/blob/3febcd7377d0cdb322bbb420207c3c26cbf7afd7/kubeflow/spark/backends/kubernetes/constants.py#L44

This conflicts with the provision in the `Driver` dataclass to pass in a custom Spark image.

https://github.com/kubeflow/sdk/blob/3febcd7377d0cdb322bbb420207c3c26cbf7afd7/kubeflow/spark/types/types.py#L89

This provision can be leveraged from the public `connect` API.

https://github.com/kubeflow/sdk/blob/3febcd7377d0cdb322bbb420207c3c26cbf7afd7/kubeflow/spark/api/spark_client.py#L69

### What did you expect to happen?

There should not have been any conflict.

Possible solutions would be to make Spark version configurable, derive it from the image or to validate them both for version mismatch.

### Environment

Kubernetes version:
```bash
$ kubectl version

```
Kubeflow Trainer version:
```bash
$ kubectl get pods -n kubeflow -l app.kubernetes.io/name=trainer -o jsonpath="{.items[*].spec.containers[*].image}"

```
Kubeflow Python SDK version:
```bash
$ pip show kubeflow

```

### Impacted by this bug?

Give it a 👍 We prioritize the issues with most 👍