What a replica computes on comes from nvidia-smi; llama.cpp says its layers only at -lv 4

Checked against the installed llama.cpp (b10991): at its default verbosity
it prints neither the devices it uses nor the layers it offloads, so the
start-up output said nothing and every replica's ON column stayed empty.
The agent now takes compute from nvidia-smi - the GPUs the shard's processes
hold memory on, and how much - or CPU, for every engine, falling back to it
when llama.cpp's output says nothing; adopted engines get it too, and each
session reports it once per shard as a compute-only status. engine-check
runs its GPU token at -lv 4, where llama.cpp does say where its layers went.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

internal/cli/enginecheck.go

@@ -363,7 +363,9 @@ func computeOnGPU(ctx context.Context, bin, model string, out io.Writer) error {
 	defer cancel()
 	var logs syncBuffer
 	c := exec.CommandContext(ctx, bin, "-m", model, "--host", "127.0.0.1", "--port", strconv.Itoa(port), //nolint:gosec // the engine's own binary
-		"--n-gpu-layers", "999", "--ctx-size", "512", "--parallel", "1")
+		"--n-gpu-layers", "999", "--ctx-size", "512", "--parallel", "1",
+		// Where the layers went is said only at this verbosity.
+		"-lv", "4")
 	c.Stdout, c.Stderr = &logs, &logs
 	if err := c.Start(); err != nil {
 		return fmt.Errorf("starting %s: %w", bin, err)