Add voxi_speak.py ([emotion] tag wrapper) + document tag synthesis

Voxi expressive Kokoro fine-tune: inference model, 8 emotion voicepacks, checkpoints

initial commit