Remove a peer's live routes when its session goes down (closes #56)

RIS sends no withdrawals for the routes of a peer whose BGP session goes
down, so routewatch kept them.

The streamer now hands STATE messages to the prefix handler. They are never
dropped for backpressure: one that finds the queue full is held and queued
first once there is room, and the handler's other messages are dropped
meanwhile, so none overtakes it. Nothing waits. On a down, the prefix
handler drops the peer's pending updates and its next flush deletes all the
peer's live routes before the batch's other writes, keeping the in-memory
counts exact. The live route tables are unique on peer and prefix, which
also finds a peer's routes.

Model: opus-5-5

internal/routewatch/peerhandler.go

@@ -1,7 +1,7 @@
 package routewatch
 
 // peerhandler.go provides batched peer tracking functionality for BGP route monitoring.
-// It tracks BGP peers from all incoming RIS messages and maintains peer state in the database.
+// It tracks BGP peers from incoming UPDATE messages and maintains peer state in the database.
 
 import (
 	"strconv"
@@ -33,7 +33,7 @@ const (
 	peerBatchTimeout = 2 * time.Second
 )
 
-// PeerHandler tracks BGP peers from all message types using batched operations.
+// PeerHandler tracks BGP peers from UPDATE messages using batched operations.
 // It maintains a queue of peer updates and periodically flushes them to the database
 // in batches to improve performance. The handler deduplicates peer updates within
 // each batch, keeping only the most recent update for each peer IP address.
@@ -82,11 +82,11 @@ func NewPeerHandler(db database.Store, logger *logger.Logger) *PeerHandler {
 	return h
 }
 
-// WantsMessage returns true for all message types since peer information
-// is extracted from every RIS message regardless of type. This satisfies
-// the MessageHandler interface.
-func (h *PeerHandler) WantsMessage(_ string) bool {
-	return true
+// WantsMessage returns true for UPDATE messages only. A peer whose session
+// never comes up sends STATE messages and nothing else, so a STATE message is
+// not a sign of a peer that carries routes.
+func (h *PeerHandler) WantsMessage(messageType string) bool {
+	return messageType == ristypes.MessageTypeUpdate
 }
 
 // QueueCapacity returns the desired queue capacity for this handler.