Avoid initial allocation delay in IPAMFromCloud and IPFromCluster allocator

[satyasm]

Is this a BUG REPORT or FEATURE REQUEST?:
/kind bug

What happened:
The onAdd handler in ipam/controller.go

	if syncer, ok := c.syncers[node.Name]; !ok {
		syncer = c.newSyncer(node.Name)
		c.syncers[node.Name] = syncer
		go syncer.Loop(nil)
	} else {
		glog.Warningf("Add for node %q that already exists", node.Name)
		syncer.Update(node)
	}

create a new sync instance with it's own loop, but does not actually handle the event itself. This means that the loop code falls back to the delay handler case (https://github.com/kubernetes/kubernetes/blob/master/pkg/controller/nodeipam/ipam/sync/sync.go#L146) in the absence of any new updates seen on this node. This might mean that in the extreme case of no new update events seen for the node after the add, it can take up to 30 seconds for the pod CIDR to be allocated. Granted that in most practical cases, during node addition, other updates would trigger an earlier allocation.

What you expected to happen:
Trigger the syncer.Update code in all cases, including node add events.

How to reproduce it (as minimally and precisely as possible):
Observed as potential delay in pod CIDR allocataion and how long node takes to be fully ready when using IPAMFromCluster or IPAMFromCloud allocators.

Anything else we need to know?:

Environment:

• Kubernetes version (use kubectl version): 1.9
• Cloud provider or hardware configuration: GCE
• OS (e.g. from /etc/os-release): all supported?
• Kernel (e.g. uname -a): all supported?
• Install tools: all supported?
• Others:

[satyasm]

/sig network