remove routing rules that became stale while glooshot was not watching

[mitchdraft]

reproduction steps:

• start glooshot
• create an experiment
• wait for glooshot to produce routing rules from this experiment
• stop glooshot
• delete the experiment
• restart glooshot
• Expect Behavior glooshot to see that there are no experiments, and thereafter update the routing rules
• Actual Behavior glooshot does not remove the stale routing rules until the experiment emitter is triggered with the creation of a new experiment

potential fix:

• call sync by hand when getting a new Syncer:

  func NewSyncer(ctx context.Context, expClient v1.ExperimentClient, rrClient sgv1.RoutingRuleClient, meshClient sgv1.MeshClient, opts options.Opts) (*glooshotSyncer, error) {
  syncer := &glooshotSyncer{
      expClient:    expClient,
      rrClient:     rrClient,
      rrReconciler: sgv1.NewRoutingRuleReconciler(rrClient),
      meshClient:   meshClient,
      opts:         opts,
  }
  // must run sync on initialization in order to remove any stale resources
  // get a new context to ensure the root context is not cancelled, timeout is arbitrary
  initCtx, _ := context.WithTimeout(ctx, 10*time.Second)
  return syncer, syncer.Sync(initCtx, &v1.ApiSnapshot{})
  }

• issues with this approach: must not pass an empty snapshot if there are existing resources

better fix

• update solokit such that the emitters take an "emit on initialization" boolean in their constructor that tells it whether or not it should force-emit on startup

[mitchdraft]

@ilackarms mentioned that Gloo has a way of dealing with this for UDS