optimization: generate children counts on demand

Accumulating counts in the Node data structure takes a lot of effort,
and it turns out we only use this information in a few places where
it's easy to compute.

report/metadata_template.go

@@ -71,8 +71,11 @@ func fromSets(n Node, key string) (string, bool) {
 }
 
 func fromCounters(n Node, key string) (string, bool) {
-	val, ok := n.LookupCounter(key)
-	return strconv.Itoa(val), ok
+	val := n.CountChildrenOfTopology(key)
+	if val == 0 {
+		return "", false
+	}
+	return strconv.Itoa(val), true
 }
 
 // MetadataRow is a row for the metadata table.

report/node.go

@@ -100,6 +100,17 @@ func (n Node) AddCounter(k string, value int) Node {
 	return n.WithLatest(name, mtime.Now(), strconv.Itoa(value))
 }
 
+// CountChildrenOfTopology returns how many children have a particular topology
+func (n Node) CountChildrenOfTopology(topology string) int {
+	count := 0
+	for _, childID := range n.ChildIDs {
+		if ty, ok := NodeIDType(childID); ok && ty == topology {
+			count++
+		}
+	}
+	return count
+}
+
 // WithSet returns a fresh copy of n, with set merged in at key.
 func (n Node) WithSet(key string, set StringSet) Node {
 	n.Sets = n.Sets.Add(key, set)