[Docs] README and guide status tables are stale; CPU/GPU claims #121

README "Status" still says model choice is hard-coded, the OpenAI-compatible API is planned and tests are "not yet"; faq.html says automated tests and continuous batching are planned. The README mentions "your own GPU" while the engine defaults to device="cpu" and CUDA is deferred, yet pyproject.toml pulls CUDA wheels on Windows/Linux.

Done when: README, README_ko and the guide match SPEC and PROJECT §2; the GPU wording and the CUDA wheel sources are reconciled (CPU wheels or a documented device option).

Found in the 2026-10-03 usability audit (code reading, not run).