Add Vevo2 source audio chunking

[0xShug0]

Adds opt-in source-audio chunking for Vevo2 voice conversion and singing voice conversion routes.

What changed:

• Adds audio_chunk_duration_sec to split long source audio for style_preserved_vc and style_preserved_svc.
• Adds cross_fade_duration_sec for chunk overlap and output crossfade.
• Keeps the existing one-shot path unchanged when chunking is not requested.
• Exposes the new controls in the native UI and documents the CLI request options.

Resolves #481.