Add preview samples for built-in preset voices

Feature request

Built-in/preset voices such as Kokoro and Qwen CustomVoice should have an easy way to hear each voice before creating/selecting a profile.

Current behavior

Preset voice pickers list names, gender, and language, but there is no preview audio. Users have to create/select a preset profile and run a generation manually to hear what the voice sounds like.

Proposed behavior

Add a preview/play button for each preset voice in the built-in voice picker and/or voice list.

Suggested implementation options:

• Generate a short default phrase on demand, e.g. "Hello, this is a quick voice test. I am checking clarity, warmth, and natural pacing."
• Cache generated previews per engine + voice id + language so the preview is not regenerated every time.
• Show loading/error states beside the voice card.
• Reuse existing generation/audio playback plumbing where possible.

Why this matters

Preset voice libraries are hard to evaluate by name alone. Preview audio would make it much faster to compare Kokoro/Qwen voices and choose the right voice before generating longer text.