kernel - Implement new callout*() core

kernel - Implement new callout*() core

* Rewrite the callout*() function core to run more efficiently.
  This has a much better mechanism for locking a callout to a
  cpu and for dealing with synchronous waits for callbacks to
  complete.

  The IPI busy/wait loop has been removed for remote-cpu operations.
  The code now blocks normally and callers will have to understand that.
  This will make it a lot easier to debug races.

* Add callout_init_lk() which implements auto-locking similar
  to FreeBSD.  Several FreeBSD mechanisms already depend on it
  and it will make porting easier.  And it works pretty well.

* Preparation for a more synchronous interface.

* Note that the new API is roughly similar to the old
  except callout_reset() now issues a synchronous stop
  instead of an asynchronous stop.  Soon we will also
  switch around the function names to make MP operation
  and synchronous operation the default across the
  board.