site: Add faq for Linkerd integration with Contour

[stevesloka]

When using Linkerd with Contour, the shutdown-manager logic isn't consistent and doesn't always work. The problem is when a SIGTERM is received, linkerd kills the network so that the preStop, http request cannot be made. This causes the Envoy pod to terminate immediately.

Linkerd has a configuration option called --wait-before-exit-seconds which can set a window of time to wait before exiting. Might be good to set this to the podTerminationGracePeriodSeconds parameter?

In any case we should update our faq to help out future users if they run into the same issue.

Slack thread: https://kubernetes.slack.com/archives/C8XRH2R4J/p1588882697463700?thread_ts=1588862631.455000&cid=C8XRH2R4J Upstream KEP: https://github.com/kubernetes/enhancements/issues/753