Merge Mark Nevill's distributed capability system.

This the implementation of work discussed in the scope of Mark Nevill's
master's thesis (http://www.barrelfish.org/nevill-master-capabilities.pdf).

The system works roughly as follows:
* Each capability has an owner which is responsible to synchronize operations
  that need synchronization in order to preserve system-wide consistency.
* The operations that need synchronization are implemented in the monitors with
  synchronization being implemented as broadcasts on the intermon bindings and
  called whenever a local capability management invocation returns with the
  error message SYS_ERR_RETRY_THROUGH_MONITOR.
* These calls are done as blocking calls to the local monitor.

Newly introduced terminology/concepts:
* Ownership: each capability has an owner that is the leader for operations
  that need synchronization.
* Move: the move operation changes the owner of a capability: currently
  allowed for capabilities of the types PhysAddr, RAM, and Frame

Known Bugs:
* Some operations that should do capability move operations in the monitors
  still transfer the bit-representation of the capability manually instead of
  doing a move.

Open Issues:
* FCNodes are not implemented yet
* VNode and DevFrame ownership cannot be moved