eb_device_close kills outstanding callbacks, not eb_socket_close.
... to do this, response tracked cycles remember their device.
... and on device close it filters the entire list of awaiting responses.

The reasoning behind this change:
  the user probably has handles to the device in the callback
  if he closes the device, the callbacks might run later with a bad pointer
  instead => kill those callbacks as part of closing the device.

These will work much nicer in hardware.

If you had no attempts on device_open, the cached time was not set.
Then your next operation would instantly time out.
Solution => set time_cache on socket open.

... encourage people to use the proper API.

... I wish more libraries did this!

Remove 8-bit eb_address_t support.
... The config space is at least 16 bit wide.

No commit message

No commit message

No commit message

Otherwise you get an 'unused' warning.

   inline
   unnamed unions
So, change all occurances for improved portability.

Work around this.

I think its safe to assume our command-line tools have such an environment.