cuda: only scale launch queues for multi-GPU systems#19230

BusinessBuilders wants to merge 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

BusinessBuilders:fix/skip-launch-queue-scaling-tegraBusinessBuilders/llama.cpp:fix/skip-launch-queue-scaling-tegraCopy head branch name to clipboard