Hubble Relay implements Hubble's `observer.ServerStatus` endpoint from
a cluster-wide perspective. It forwards `ServerStatus` requests to
Hubble peers in the cluster and aggregates the responses. However, while
the total number of seen flows can be accumulated, the uptime value
cannot. Therefore, the intent for Relay's implementation of
`ServerStatus` regarding the uptime value was to return the uptime of
the oldest running Hubble server in the cluster. However, the
implementation is wrong and Relay reports the newest uptime from all the
nodes in the cluster.

In both the Hubble CLI and Hubble UI, the flow rate is reported as the
number of flows per second, e.g. when `hubble status`, which is computed
as the number of seen flows divided by the uptime. As you can imagine,
the flows/s rate would be very wrong when a Hubble server is restarted
as the accumulated seen flows in the cluster would be a large number
while the uptime, because of the bug mentioned above, would be a very
small number resulting in wrong reporting of the flows rate.

This patch fixes this bug by correctly reporting the oldest uptime from
all connected Hubble peers instead of the newest one.

Signed-off-by: Robin Hahling <robin.hahling@gw-computing.net>