Ignore networks that disappear during discovery

[jamie01]

On systems where we have multiple concurrent deployments and teardowns of virtual machines that overlap, there are occasions where setup can fail if teardown of another vagrant deployment is happening in parallel. Generally this is handled well, but the most common place of failure was in the network enumeration where details of all existing networks are collected prior to creating new ones.

Since a number of operations are done after driver.list_all_networks and in cases where the number of networks is relatively large, it was common for a network to be destroyed during the iterations of that loop and thus the XML lookup or queries for bridge_name etc would fail.

My change here is just to rescue any errors that occur in this loop and continue on. I understand this is a bit of a niche issue for us in the way vagrant-libvirt is used at scale, and this might mask other problems besides networks being cleaned up, but opening this PR anyway for thoughts.