ocserv autoscale needs latency to determine server instance health

A key measure of health of an ocserv instance is the latency it is experiencing when processing packets. Ideally we would want to measure both inbound and outbound latency, but the Linux tun adapter currently drops RX stats from the packet.

As a first pass, ocserv should measure the time from when the packet arrives on the network adapter until it is transmitted out the tun adapter. Once the tun adapter is fixed to pass inbound rx timestamps up, we can measure the reverse as well.

Latency is the preferred metric to quantify instance health as it directly impacts the end user experience.