feat(cluster): add probe execution mode, check extraction, and result aggregation

Phase 2 of distributed probing:
- Extract check logic into standalone RunCheck() for use by probes
- Add probe cluster mode: stateless nodes that fetch assignments, execute
  checks, and report results to the leader
- Add multi-node result aggregation with configurable strategy
  (any-down, majority-down, all-down)
- Leader ingests probe results into engine live state and triggers alerts
- New env vars: UPKEEP_NODE_ID, UPKEEP_NODE_NAME, UPKEEP_NODE_REGION,
  UPKEEP_AGG_STRATEGY
- Example docker-compose.probe.yml with leader + 2 regional probes

internal/monitor/aggregator.go

@@ -0,0 +1,44 @@
+package monitor
+
+import "time"
+
+type AggregationStrategy string
+
+const (
+	AggAnyDown      AggregationStrategy = "any-down"
+	AggMajorityDown AggregationStrategy = "majority-down"
+	AggAllDown      AggregationStrategy = "all-down"
+)
+
+type NodeResult struct {
+	NodeID    string
+	IsUp      bool
+	LatencyNs int64
+	CheckedAt time.Time
+}
+
+func AggregateStatus(results []NodeResult, strategy AggregationStrategy) (isUp bool, avgLatencyNs int64) {
+	if len(results) == 0 {
+		return true, 0
+	}
+
+	upCount := 0
+	var totalLatency int64
+	for _, r := range results {
+		if r.IsUp {
+			upCount++
+		}
+		totalLatency += r.LatencyNs
+	}
+	avgLatencyNs = totalLatency / int64(len(results))
+
+	switch strategy {
+	case AggMajorityDown:
+		isUp = upCount > len(results)/2
+	case AggAllDown:
+		isUp = upCount > 0
+	default:
+		isUp = upCount == len(results)
+	}
+	return
+}