Add explicit cleanup for Docker network monitor autoruns

The DNS server and tunnel container connections weren't cleaned up.
The existing dispose stopped any further events starting, which
helps, but pending events doing async work (getting network aliases)
could have resulted in changes after the monitor was shutdown. Disposing
all connections avoids this.

Author: pimterry

src/interceptors/docker/docker-networking.ts

@@ -104,15 +104,17 @@ export async function monitorDockerNetworkAliases(proxyPort: number): Promise<Do
         const networkMonitor = new DockerNetworkMonitor(docker, proxyPort, stream);
 
         // We update DNS immediately, and on all changes:
-        mobx.autorun(() => dnsServer.setHosts(networkMonitor.aliasIpMap));
+        const dnsDisposer = mobx.autorun(() => dnsServer.setHosts(networkMonitor.aliasIpMap));
+        networkMonitor.addDisposer(dnsDisposer);
 
         // We update tunnel _only_ once something is actually intercepted - once interceptedNetworks changes.
         // We don't want to create the tunnel container unless Docker interception is actually used.
-        mobx.reaction(
+        const tunnelDisposer = mobx.reaction(
             () => networkMonitor.interceptedNetworks,
             (interceptedNetworks) => updateDockerTunnelledNetworks(proxyPort, interceptedNetworks)
                 .catch(console.warn)
         );
+        networkMonitor.addDisposer(tunnelDisposer);
 
         dockerNetworkMonitors[proxyPort] = networkMonitor;
 
@@ -194,8 +196,15 @@ class DockerNetworkMonitor {
         this.refreshAllNetworks();
     }
 
+    private disposers: Array<() => void> = [];
+
+    addDisposer(disposer: () => void) {
+        this.disposers.push(disposer);
+    }
+
     async stop() {
         this.dockerEventStream.removeListener('data', this.onEvent);
+        this.disposers.forEach(d => d());
     }
 
     private readonly networkTargets: {