TTS fails unless a voice is specified on the command line.

As reported by Stepan Novotny on https://bugs.debian.org/1101681

### Steps to reproduce

`spd-say -o espeak-ng-mbrola 'Testing testing 1 2 3'`

### Obtained behavior

This fails to use the default (mbrola) voice

### Expected behavior

This should select an mbrola voice (according to language etc.)

### Behavior information

Running `spd-say -o espeak-ng-mbrola -y mb-us1 'Testing testing 1 2 3'` does get an mbrola voice.

I tried adding voice configuraiton to /etc/speech-dispatcher/modules/espeak-ng-mbrola.conf

```
VoiceFileDependency "/usr/share/mbrola/$VOICE/$VOICE"
VoiceFileDependency "/usr/lib/x86_64-linux-gnu/espeak-ng-data/voices/mb/mb-$VOICE"
AddVoice "en-us" "FEMALE1" "us1"
AddVoice "en-us" "MALE1" "us2"
DefaultVoice "us1"
```

The above paramters are all ignored.

### Distribution

Debian 13

### Version of Speech-dispatcher

0.12.0