metrics scrape_job for coordinator's nginx is misconfigured when num_units > 1

### Bug Description

Unconfirmed, but I suspect we only emit nginx metrics from the Coordinator's leader unit when Coordinator is scaled. I see Coordinator instantiates a metrics relation with a scrape job for all workers [and this particular coordinator unit](https://github.com/canonical/cos-coordinated-workers/blob/245d0c23922c38e76590248f5d30574d67d624cb/src/coordinated_workers/coordinator.py#L651-L659). If a coordinator is scaled, all units will independently create this same job (with their own unit's nginx) but **only the leader will send that data because MetricsEndpointProvider only sends data from the leader**. So what I think we'll see is nginx logs from only the leader's unit and no others

### To Reproduce

deploy with coordinator having scale=2 and scrape metrics

### Environment

.

### Relevant log output

```shell
.
```

### Additional context

_No response_

Start with src/coordinated_workers/coordinator.py at the metrics relation and scrape-job setup around lines 651-659, then inspect how MetricsEndpointProvider handles data from scaled units. Reproduce the issue with Coordinator scaled to two units and scrape the metrics. Done means nginx metrics are emitted for every coordinator unit rather than only the leader.

Written by the indexing model from the issue text.