Update dataset card with train/eval/test split info

Replace single split with train/eval/test splits (20300/1001/120 samples)

Update README.md

Upload cleaned Georgian TTS dataset (21,421 samples, 35h)

initial commit