Initial commit: full Phases 1-6 implementation

Post-repair hardware validation pipeline for Proxmox cluster hosts.
Go orchestrator + in-image agent + mkosi live image + bundled dnsmasq
PXE + SQLite + HTMX/SSE UI + notify registry + janitor + full docs.

agent/tests/stage.go

@@ -0,0 +1,67 @@
+// Package tests contains the per-stage executors the agent runs on the
+// host under test. Each stage implements Runner, is called with a
+// Context that carries the client + forwarder + run params, and returns
+// an Outcome that the caller POSTs to /result.
+package tests
+
+import (
+	"context"
+	"encoding/json"
+	"time"
+)
+
+// Outcome is what a stage returns; it maps directly to the /result body.
+//   - Passed=true and len(Skipped)>0 counts as a pass but surfaces in the
+//     tile summary so operators can see "GPU: skipped (no VGA device)".
+//   - Message is only used on failure; the UI displays it in the log.
+//   - Extras is merged into the posted summary so stages can add
+//     their own shape (e.g. Storage returns per-disk probe results).
+type Outcome struct {
+	Passed  bool
+	Message string
+	Summary string         // short human-readable one-liner
+	Extras  map[string]any // merged into posted summary JSON
+}
+
+// MarshalSummary builds the summary JSON body POSTed to /result.
+// Stages accumulate fields via Extras; this helper adds "summary" (the
+// human-readable line) and serializes.
+func (o Outcome) MarshalSummary() (json.RawMessage, error) {
+	body := map[string]any{}
+	for k, v := range o.Extras {
+		body[k] = v
+	}
+	if o.Summary != "" {
+		body["summary"] = o.Summary
+	}
+	return json.Marshal(body)
+}
+
+// Deps bundles what stages need without pulling in the whole agent.
+// Logger methods print to stdout + forward to the orchestrator; Sensor
+// drops numeric samples; OverrideFlags carries operator-set bypasses.
+type Deps struct {
+	Info           func(string)
+	Warn           func(string)
+	Error          func(string)
+	Sensor         func(ctx context.Context, samples []Sample) error
+	OverrideWipe   bool
+	ExpectedDisks  []ExpectedDisk // serials + sizes from host.expected_spec
+	StageTimeout   time.Duration
+}
+
+// Sample mirrors the server's SensorSample but lives in the tests
+// package so probe code doesn't import internal/api.
+type Sample struct {
+	Kind  string
+	Key   string
+	Value float64
+	Unit  string
+}
+
+// ExpectedDisk is the subset of internal/spec.DiskSpec that Storage
+// needs: a device allowlist keyed on serial.
+type ExpectedDisk struct {
+	Serial string
+	SizeGB int
+}