Log model backends at startup (or surface in a status endpoint)

## Problem
A developer enables `scope.models.embed()` in their resource code, ships, and `embed()` returns... something? Maybe an error? It's not obvious which backend is wired up, which model is the default, or whether the configuration was even read.

Today there's no startup log line that says \"here's what models you've got.\" If you misspelled `defaultBackend`, you find out the first time `embed()` runs.

## Proposal
At startup (after the existing \"Harper Pro successfully started\" banner), emit:

```
Model backends:
embed: ollama @ localhost:11434
default model: nomic-embed-text
generate:
```

When no backend is configured at all (and no ollama auto-detect from the related issue), log a single info line:

```
Model backends: none configured. scope.models.embed() will throw ModelBackendNotFoundError.
```

Behind `logger.info`, so it doesn't spam debug-level logs. The point is making the configured state grep-able from `docker logs` without reading code.

## Stretch: status endpoint
Add `models` to the `harper get_status` / `system_information` operation output:

```json
{
\"models\": {
\"embed\": { \"backend\": \"ollama\", \"host\": \"localhost:11434\", \"defaultModel\": \"nomic-embed-text\" },
\"generate\": null
}
}
```

Useful for ops automation and the Fabric UI.

## Related
- PR #779 (Phase 1 model analytics emission)