libhal-policy -> PolicyKit

[Artem Kachitchkine]

> I thought about this and I think we are safe:
> 
>  1. The PolicyKit daemon will be single threaded and process requests
>     sequentially; and
> 
>  2. temporary_policy_override will just be an in-memory object - the
>     backing store will _not_ be modified for this; and
> 
>  3. The Disconnected signal from the message bus for pid X
>     is guaranteed to arrive before a new call to the PolicyKit service
>     checking for the stuff in temporary_policy_override is done by
>     another pid X.

As long as the daemon's request queue is an in-memory object, I think 
we're safe even without the two last items. In any case I wouldn't rely 
on 3.

-Artem.