v3 slim: target checkpoint + fine-tuned vocoder + ONNX (base checkpoint withdrawn)

v3: model card quality table + samples/eval.md

v3: new data recipe, 18 style tokens, fine-tuned HiFi-GAN vocoder, base checkpoint for fine-tuning

rename to Matcha-TTS-PL (README.md)

rename repos: drop -v2 suffix (README.md)

move playground to github.com/machinekind/tts-pl-playground; add baked-voices ONNX (v2)

release tts-pl-matcha-v2

initial commit