ui: GitHub-Actions-style detail page, sub-steps, mini-tile run-view

Reshapes the detail page into a run-view: hybrid horizontal pipeline
+ expanded active-step pane with sub-steps, a per-step log pane with
line-numbered permalinks and client-side search, and a runs-history
sidebar that navigates via ?run=N. Default step is server-picked
(running → failed → Reporting) so the operator lands on the thing
that's moving.

Adds a sub_steps table + SSE topic (substep-{run}-{stage}-{ordinal})
so per-disk and per-pass work (SMART, CPUStress CPU/RAM, Storage,
GPU) is visible in the UI instead of buried in stage summary JSON.
Agent emits sub-step reports from existing per-iteration loops.

Dashboard tiles become a mini run-view with a 9-dot step strip so
the operator reads run health across the whole grid at a glance.
Register page gets the same card shell + button styling.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

agent/tests/gpu.go

@@ -2,8 +2,11 @@ package tests
 
 import (
 	"context"
+	"encoding/json"
+	"fmt"
 	"os/exec"
 	"strings"
+	"time"
 )
 
 // GPU enumerates VGA / 3D PCI devices. No devices → skip cleanly (a
@@ -11,7 +14,9 @@ import (
 // stress). Devices present → try nvidia-smi for NVIDIA cards, else
 // accept PCI presence.
 func GPU(ctx context.Context, d Deps) Outcome {
+	pciStart := time.Now()
 	devices := listGPUPCI(ctx)
+	pciEnd := time.Now()
 	if len(devices) == 0 {
 		d.Info("GPU: no VGA/3D PCI devices found — skipping stage")
 		return Outcome{
@@ -22,7 +27,9 @@ func GPU(ctx context.Context, d Deps) Outcome {
 	}
 	d.Info("GPU: found " + joinDevices(devices))
 
+	nvStart := time.Now()
 	nvidia := nvidiaSmiList(ctx)
+	nvEnd := time.Now()
 	extras := map[string]any{
 		"pci_devices": devices,
 		"skipped":     false,
@@ -31,10 +38,39 @@ func GPU(ctx context.Context, d Deps) Outcome {
 		extras["nvidia"] = nvidia
 		d.Info("GPU: nvidia-smi reports: " + strings.Join(nvidia, ", "))
 	}
+
+	// Sub-step rows: one per enumerated PCI device, plus (optionally) one
+	// per NVIDIA card when nvidia-smi sees anything. PCI enumeration runs
+	// once for all devices — we bracket that single invocation by
+	// pciStart/pciEnd and attribute the window to each device row so the
+	// UI can still slice the log per row by time.
+	var subs []SubStepReport
+	for i, dev := range devices {
+		summary, _ := json.Marshal(map[string]any{"pci": dev, "ordinal": i})
+		subs = append(subs, SubStepReport{
+			Name:        fmt.Sprintf("pci #%d", i),
+			Passed:      true,
+			StartedAt:   pciStart,
+			CompletedAt: pciEnd,
+			SummaryJSON: summary,
+		})
+	}
+	for i, line := range nvidia {
+		summary, _ := json.Marshal(map[string]any{"nvidia_smi": line})
+		subs = append(subs, SubStepReport{
+			Name:        fmt.Sprintf("nvidia #%d", i),
+			Passed:      true,
+			StartedAt:   nvStart,
+			CompletedAt: nvEnd,
+			SummaryJSON: summary,
+		})
+	}
+
 	return Outcome{
-		Passed:  true,
-		Summary: formatCount(len(devices), "GPU present"),
-		Extras:  extras,
+		Passed:   true,
+		Summary:  formatCount(len(devices), "GPU present"),
+		Extras:   extras,
+		SubSteps: subs,
 	}
 }