Add TADA speech engine integration

[diodiogod]

Summary

• add Hume AI TADA 1B and TADA 3B multilingual TTS support
• integrate unified Text, SRT, character switching, caching, progress, interruption, and model lifecycle handling
• add selective model/tokenizer downloads, shared Transformers 4 runtime support, compilation caching, documentation, and workflow cover
• require exact reference transcripts instead of silently falling back when character transcription is missing

Why

Adds issue #269 as a native engine integration while following the suite's current engine architecture and download conventions.

Validation

• manually generated working English and Portuguese audio in ComfyUI
• exercised character switching, model downloads, shared runtime loading, progress reporting, and corrected audio decoding
• verified missing reference transcripts return an actionable error

Still pending before release

• broader multilingual listening tests
• final integration workflow JSON
• final smoke tests for SRT cancellation, cache reuse, Clear VRAM/reload, and offline/manual model layout

Closes #269