Show route churn as a whole count per minute (closes #160)

The status page showed route churn per second as a one-minute moving
average, so a quiet feed gave a fraction such as 0.1. The average was
correct.

The metrics tracker now also counts prefix announcements and
withdrawals, from all peers, in each clock minute. /api/v1/stats
reports the count for the last full minute as route_churn_per_min in
stream, in place of route_churn_per_sec. The status page shows it as
"Route Churn/min".

The status and home pages round messages and updates per second to
whole numbers.

Model: opus-5-5

README.md

@@ -82,8 +82,8 @@ Every route answers `HEAD` as it answers `GET`, without the body.
 
 - `GET /api/v1/stats` - The `/status.json` figures plus connection, garbage
   collector, stream and handler queue statistics. `stream.route_churn_per_min`
-  counts the prefixes announced or withdrawn in the last full minute of the
-  clock.
+  is the number of prefix announcements and withdrawals, from all peers, in the
+  last full clock minute.
 - `GET /api/v1/ip/{ip}` - Looks up each address: its most specific live route
   (`netblock`), that route's origin AS with its WHOIS details, and the address's
   PTR names. `{ip}` is an IP address, a host name, or a comma-separated list of

TODO.md

@@ -33,11 +33,11 @@ tested.
 
 # Completed Steps
 
-- 2026-10-26: the status page shows "Route Churn/min", the whole number of
-  prefixes announced or withdrawn in the last full minute. `/api/v1/stats`
-  reports it as `route_churn_per_min` in `stream`, which replaces the moving
-  average `route_churn_per_sec`. The messages and updates per second on `/` and
-  `/status` are whole numbers (closes #160)
+- 2026-10-26: the status page shows "Route Churn/min", the number of prefix
+  announcements and withdrawals, from all peers, in the last full clock minute.
+  `/api/v1/stats` reports it as `route_churn_per_min` in `stream`, which
+  replaces the moving average `route_churn_per_sec`. The messages and updates
+  per second on `/` and `/status` are whole numbers (closes #160)
 - 2026-10-26: the WHOIS fetcher backs off more slowly. A failure makes its wait
   1.5 times as long, up to one minute instead of five. Its queries to RIPE,
   APNIC and AFRINIC leave out person and role objects, which RIPE counts against

internal/metrics/metrics.go

@@ -34,7 +34,7 @@ type Tracker struct {
 	// Announcement/withdrawal metrics
 	announcementCounter metrics.Counter
 	withdrawalCounter   metrics.Counter
-	churnPerMinute      minuteCounter // announced and withdrawn prefixes
+	churnPerMinute      minuteCounter // prefix announcements and withdrawals
 
 	// BGP peer tracking
 	bgpPeerCount atomic.Int32
@@ -198,8 +198,8 @@ func (t *Tracker) GetWithdrawalCount() uint64 {
 	return uint64(count)
 }
 
-// GetChurnPerMinute returns the number of prefixes announced or withdrawn in
-// the last full minute of the clock.
+// GetChurnPerMinute returns the number of prefix announcements and
+// withdrawals, from all peers, in the last full clock minute.
 func (t *Tracker) GetChurnPerMinute() int {
 	return t.churnPerMinute.lastFullMinute()
 }

internal/metrics/minutecounter.go

@@ -5,8 +5,8 @@ import (
 	"time"
 )
 
-// minuteCounter counts events in each minute of the clock and reports the
-// count of the last full minute. It is safe for concurrent use.
+// minuteCounter counts events in each clock minute and reports the count of
+// the last full minute. It is safe for concurrent use.
 type minuteCounter struct {
 	mu  sync.Mutex
 	now func() time.Time

internal/metrics/minutecounter_internal_test.go

@@ -44,6 +44,12 @@ func TestChurnPerMinuteCountsLastFullMinute(t *testing.T) {
 	now = time.Date(2026, 10, 26, 12, 4, 0, 0, time.UTC)
 
 	expectChurn(0, "12:03 had none")
+
+	tracker.RecordWithdrawal()
+
+	now = time.Date(2026, 10, 26, 12, 6, 0, 0, time.UTC)
+
+	expectChurn(0, "12:05 had none, though 12:04 had one withdrawal")
 }
 
 // TestChurnPerMinuteCountsConcurrentRecords announces prefixes from several