Use event based connection failure handling if libvirt is new enough

[Russell Bryant]

This is related to https://code.launchpad.net/bugs/1154473

This bug pointed out the fact that the reconnect handling used in the libvirt driver is fragile and doomed to break again in the future. To quote danpb from https://review.openstack.org/#/c/24323

"This approach is really doomed to failure in general since libvirt can change which error code is raised in a particular scenario at any release. This change isn't making life worse though, so I'll ACK it.

Libvirt >= 0.10.0 has support for triggering a formal event callback when a connection is lost. For the future, we should plan to use that in preference to this error check hack if we have new enough libvirt. "

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/32215

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/32215
Committed: http://github.com/openstack/nova/commit/24df413e4d9fea1dbc45cdbabc30475405b40bfe
Submitter: Jenkins
Branch: master

commit 24df413e4d9fea1dbc45cdbabc30475405b40bfe
Author: Jason Dillaman <email address hidden>
Date: Fri Jun 7 19:18:27 2013 -0400

    Register libvirt driver with closed connection callback

    libvirt v1.0.1 and later support a callback to alert clients
    when the connection is lost. The libvirt driver will now
    register for this callback when available so that the driver
    can properly reconnect.

    Fixes: bug #1154635
    Change-Id: Ie5f1b69cf6f0655022148386cd2be9955d37b467