If there is a loadbalancer that sits between NSX and neutron usually it has
a default http_timeout on long lived http_connections. This can cause
connections that have not been used from neutron to nsx to be reset after
sometime. This patch makes the amount time we wait to reconnect the connection
configurable so one can set it to match the balancer timeout to avoid having
to try the connection if the loadbalaner has already timed it out.
closes-bug #1382000
Change-Id: Ifca76ca6c24a2002b46b1935fbb4fc63f15c5af1
0b0fe134c3