Add pilot_005 dataset outputs

Add ASR local hardening, rejected-eval extraction, and design gaps register

Add pilot_002 outputs and pending pipeline hardening changes

feat: complete LLM CC chunking pipeline with pilot results

Add scratch training notes and metadata

chore: sync all local changes before cleanup

Add full dataset recovery artifacts and run notes

Delete data/sinhala_tts_batch03_custom/vocab.txt with huggingface_hub

Delete data/sinhala_tts_batch03_custom/raw.arrow with huggingface_hub

Delete data/sinhala_tts_batch03_custom/duration.json with huggingface_hub

Add safe finetune launcher and latest training artifacts (HF-compatible)

Add CC artifacts and training outputs for 10-video run

Add F5-TTS fine-tuning script with Sinhala vocab generation, dataset prep, training, and inference