Feature request: native Italian preset voices (male + female)

Context

We use Qwen3-TTS for an Italian on-premise platform. The service requires Italian speech with a native Italian accent.

Problem

None of the 9 preset voices is native Italian. The model card suggests vivian / serena for Italian ("good for Italian"), but they are native Chinese voices: they can synthesize Italian text, yet they retain a Chinese accent in Italian speech, which is not acceptable for our use case (professional-grade content).

Request

Add at least two native Italian preset voices (one male, one female) trained on Italian-accented data, like ryan/aiden for English.

Why this matters

• WER on Italian is good (0.948, 1.7B) but WER does not measure accent — the accent is the actual blocker (cf. #134, #323).
• A voice actor-based clone (Base variant) keeps the timbre but not the target accent (#134), and raises voice-licensing/consent concerns.

Acceptance criteria

• Preset voices whose Italian output has native Italian prosody and accent.
• Ideal: same voice also usable for EN/FR intra-sentence terms without accent drift.