Update prompt_emb_en.pt from new en_ref.wav voice

Update lora prompt embed — new 3.96s/24kHz ref matching training transcript register

Replace lora prompt embed — trimmed ref 4.23s (was 11.92s, OOD for vocoder)

Add lora (UK female) prompt embedding — extracted from 11.92s ref

include lscodec/distributed/ (needed by quantization.core_vq)

LSCodec inference bundle: vocoder + codebook + pre-computed UK/EN prompt embeds

initial commit