MQ4 re-quant from FP safetensors + embedded official chat_template (1783B); current-branch quantizer, serve-verified

registry overhaul: v3-awq-f1 / q8 / mq4-moe

Add lfm2.5-1.2b-thinking.q8 (hipfire Q8, validated on gfx1100)

Add README card + LFM Open License v1.0

initial commit