MPI runLauncherAsNode allocates an extra worker when numNodes is 1

### What happened?

With runLauncherAsNode: true, EnforceMLPolicy uses max(numNodes-1, 1) for the node PodSet count. For numNodes: 1, this results in one launcher plus one worker, so the requested node count becomes 2. Should numNodes: 1 produce a launcher-only topology, or should it be rejected at admission with a minimum of 2?

### What did you expect to happen?

When numNodes: 1 and runLauncherAsNode: true, the launcher should count as the single node, so the worker count should be 0 - not forced to be a minimum of 1

### Environment

Kubernetes version:
```bash
$ kubectl version

```
Kubeflow Trainer version:
```bash
$ kubectl get pods -n kubeflow-system -l app.kubernetes.io/name=kubeflow-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 👍

Start at EnforceMLPolicy, focusing on the runLauncherAsNode handling that calculates the node PodSet count. Determine the intended admission behavior for numNodes: 1, then verify that the resulting topology has one launcher and zero workers without increasing the requested node count.

Written by the indexing model from the issue text.