Make remove_network_from_dhcp_agent more robust

[Eugene Nikanorov]

In case network failover is turned on, remove_network_from_dhcp_agent method is a matter of concurrent execution by different neutron servers.
It then may throw unexpected DB exception, breaking network failover looping call.

Neet to make both remove_network_from_dhcp_agent and remove_networks_from_down_agents more robust by catching broader exception in the loop, and using safe DB delete method.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/152576

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/152576
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=ef7e6a8553954aeb4444c90f65ac3fece8be201a
Submitter: Jenkins
Branch: master

commit ef7e6a8553954aeb4444c90f65ac3fece8be201a
Author: Eugene Nikanorov <email address hidden>
Date: Tue Feb 3 17:47:46 2015 +0300

    Improve robustness of network failover

    In some cases when remove_network_from_dhcp_agent is called
    concurrently by more than one neutron server it leads to unexpected
    DB exception when NetworkDhcpAgentBinding is removed.
    Need to avoid such case and make remove_networks_from_down_agents
    method resistant to such issues by catching broad exception.

    Change-Id: I653e200d89eb795ded742bb49420f09f66625587
    Closes-Bug: #1417629