Restore stays "ongoing" forever, with no log, when it fails before the restore command (download, decryption, archive preparation) #135

Describe the bug
In execute_restore (src/services/restore/executor.rs), the download and the archive preparation use ?:

let downloaded = self.download_backup(...).await?;
let backup_file = self.prepare_archive(...).await?;

If one of them fails, the function returns before send_result is called. The caller (src/services/restore/dispatcher.rs) only logs the error in the agent container (error!("Restore failed: {}", e)); nothing is sent to the server. The same applies to TempDir::new()?.

As a result:

• the restoration stays ongoing in the dashboard forever, with no log at all;
• the server then refuses any new restore of this database: A restoration is already waiting or ongoing for this database (409), and there seems to be no way to unblock it from the dashboard.

run_restore itself is fine: errors from pg_restore / sqlpackage are turned into a failed result with their logs.

To Reproduce

1. Start a restore from a backup whose file cannot be downloaded or decrypted (for example storage unreachable from the agent, or a wrong master key).
2. The agent container logs Restore failed: ....
3. The restoration stays ongoing in the dashboard, without logs, and a new restore of the same database returns 409.

Expected behavior
Any error during the restore job is reported to the server with status failed and the logs collected so far (Download failed: ... is already logged by download_backup), so the dashboard shows the cause and a new restore can be started.

Suggested fix
In execute_restore, catch the errors of the steps before run_restore and send a failed result instead of returning early, as execute_restic_restore already does with its outcome match.

Environment

• Agent 1.25.2, server 1.35.2, Kubernetes (AKS).
• Found while reading the code for Show backup and restore logs live in the dashboard while the job is running portabase#581 (closed as a duplicate of Better logging/progress indication on backup tasks portabase#388, which is about showing logs while a job is running).