NVIDIA GPU memory-pressure sampler. Aggregates VRAM usage across every GPU on the host via nvidia-smi --query-gpu=memory.used, memory.total --format=csv,noheader,nounits.

Returns {TotalUsedBytes, TotalCapacityBytes} summed over all GPUs. A host with no GPU or where nvidia-smi is missing reports {0, 1} so the scheduler treats it as zero pressure.