In the event of a backend BIND server being unreachable for any reason,
rndc commands will persist for a very long time and can consume
significant resources. This can be seen when running devstack with
a pool configured to point at a bind server that doesn't exist - the
rndc process count can climb into the thousands.
An optional timeout has been added to rndc to alleviate this.
Change-Id: Idd61e79715b21fdd3249136cf68a7b9d3069c3f9
Related-Bug: 1896783