make : add optional CUDA_NATIVE_ARCH (ggml-org#2482)

Use the environment variable `CUDA_NATIVE_ARCH` if present to set NVCC arch. Otherwise, use `native`.

readme : update hot topics

MPT : support GQA for replit-code-v1.5 (ggml-org#3627)

Add MPT model to supported models in README.md (ggml-org#3574)