Fix VLM backbone path: local /data/taeyoung/... -> RLWRLD/RLDX-1-VLM (align with other checkpoints)

Tighten intro: compact first sentence, drop em-dash clause, reword

Restore name RLDX-1-IMG -> RLDX-1-PT-IMG

Reword: very small -> minimal drop

Document image-input difference vs RLDX-1-PT (single frame, lighter/faster)

Rename references RLDX-1-PT-IMG -> RLDX-1-IMG

Clean to release layout: add README/LICENSE/figures, drop training metadata

Import checkpoint from kimtaey/rldx-1-image

initial commit