fix(ssh): close zombie local port forwarder on SSH keep-alive ping failures #1374

Problem

In SshTunnelManager (lib/core/security/ssh_tunnel_manager.dart, lines 393–404):

Timer? keepAliveTimer;
if (config.keepAliveIntervalSeconds > 0) {
  keepAliveTimer = Timer.periodic(
    Duration(seconds: config.keepAliveIntervalSeconds),
    (_) {
      if (!client.isClosed) {
        client.ping().catchError((_) {});
      }
    },
  );
}

1. Suppressed Ping Failures:
   When an SSH bastion drops (network blip, Wi-Fi reconnection, laptop sleep/wake, firewall state timeout), client.ping() fails, but .catchError((_) {}) suppresses the error without taking action.
2. Zombie Local Forwarder:
   serverSocket continues listening on 127.0.0.1:localPort.
   When database drivers attempt to connect to 127.0.0.1:localPort, the local TCP handshake succeeds, but line 383 (client.forwardLocal(remoteHost, remotePort)) immediately fails and destroys the socket with clientSocket.destroy().
   The driver observes an immediate RST/EOF.
   Furthermore, until someone explicitly calls openTunnel again with the same pool key, the dead session remains in _sessions[poolKey].

Scope

• In keepAliveTimer, track consecutive ping failures (e.g. 2–3 failed pings).
• When pings fail consecutively, proactively close serverSocket and client, removing the session from _sessions.
• Ensure new connection requests to that pool key dial a fresh SSH tunnel rather than reusing a zombie session.
• Add unit tests simulating ping failure and verifying proactive session cleanup.