2 - Log PacketIns with no payload and stop processing
3 - Remove unused IOFController interface

This reverts commit 51ca40a3.

Conflicts:

	src/main/java/net/floodlightcontroller/devicemanager/internal/DeviceManagerImpl.java

Now we will at least get an error message when an Exception occurs
instead of silently killing the thread.
[Forgot to add all files to commit]

Now we will at least get an error message when an Exception occurs
instead of silently killing the thread.

/wm/staticflowentrypusher/list/{switch}/json
{switch} can be 'all' or a DPID in Hex format, i.e. '00:00:00:00:00:00:00:01'

Make static flow pusher into a service (IStaticFlowEntryPusherService). This will allow other Floodlight modules to push static flows without using the REST API.

Add 5 seconds for attachment point change from one broadcast domain to another.  When the attachment points flap quickly, the controller should avoid learing the attachment point but should continue processing the packet.

Device manager will have a 5-second time gap to change an attachment point for a host from one broadcast domain to another.

Introduce a convenience method inSameIsland, which is the same as inSameCluster.  In future, they may have different functionalities.

Introduce a method called isConsistent(..) that compares an old attachment and a new attachment point. If the new attachment point is not consistent with the old one, then it has to be learnt by the device manager.

Fix a bug for checking BDDP packet if the packet is not unicast for it to continue down the processing chain.

If the incoming node port is blocked, allow only unicast and BDDP packets to continue down the processing chain.

When the device manager gets cleaned up (or when during topologyChanged), it should consider the time difference between switching from broadcast to non-broadcast attachment points (as they might be previously on two different switches).