Function run_gpu_capability_campaign 

pub fn run_gpu_capability_campaign(
    model: Option<&Path>,
    out_dir: Option<&Path>,
    tokens: u32,
) -> Result<(), String>

Probe which native GPU tiers this host has, measure the backend × mode decode matrix in child processes, and write machine-gpu-profile.json + apply-machine-gpu.ps1.

WGSL/wgpu is the portable floor; CUDA-C/PTX, HLSL+DXC, MSL, subgroups and coopmat are higher tiers when present. Only a coherent measurement can be recommended (speed without sense is failure), so the profile ranks by tok/s among coherent rows only.

One child process per cell: shared_gpu is process-wide, so QUALIA_WGPU_BACKEND cannot be re-pointed in-process.