Increase delay between pinging bootnodes (#4490)

Ensures that the delay between pings is longer than the session timeout from discovery so previously failed ping attempts are forgotten and we retry from a clean slate. This is a quick fix, ultimately the session management in discovery needs to be adjusted to drop sessions that are non-responsive faster when they are in the initial handshake phase.

CHANGELOG.md

@@ -34,3 +34,4 @@ For information on changes in released versions of Teku, see the [releases page]
  - Set an idle timeout for metrics connections, to clean up ports when no longer used.
  - Fixed error when building from a source download rather than a git checkout.
    Now logs a warning when building without git information to warn users that version information will not be available.
+ - Fixed an issue where discovery did not recover if it was initially started while the bootnodes were unavailable.

networking/p2p/src/main/java/tech/pegasys/teku/networking/p2p/discovery/discv5/DiscV5Service.java

@@ -51,7 +51,9 @@
 public class DiscV5Service extends Service implements DiscoveryService {
   private static final Logger LOG = LogManager.getLogger();
   private static final String SEQ_NO_STORE_KEY = "local-enr-seqno";
-  private static final Duration BOOTNODE_REFRESH_DELAY = Duration.ofMinutes(2);
+  // Currently must be longer than the 3 minute NodeSession timeout in discovery
+  // See https://github.com/ConsenSys/teku/pull/4490
+  private static final Duration BOOTNODE_REFRESH_DELAY = Duration.ofMinutes(4);
   private final AsyncRunner asyncRunner;
   private final Bytes localNodePrivateKey;
   private final SchemaDefinitionsSupplier currentSchemaDefinitionsSupplier;