Envoy PreStop hook support

[achevuru]

Issue #, if available:
#288
Description of changes:

Support for a configurable pre-stop delay for Envoy containers.

• Will help with keeping envoy up and running until the app container goes down. Sidecar feature in K8S 1.18 will address this natively.
• Delay also helps to account for the time it takes for the pod down event to propagate to the remaining envoys in the Mesh.
• Default value of 20 sec is based on some internal tests and specifically tuned for CloudMap based ServiceDiscovery mode and can be adjusted higher based on the scale. Value can be adjusted lower for DNS scenarios.
• If preStop delay is set to a value greater than 30 secs, then terminationGracePeriodSeconds in the pod spec (spec.terminationGracePeriodSeconds) has to be adjusted accordingly. If not adjusted, Pod will be killed(via SIGKILL) prior to the completion of envoy preStop hook as the default value of terminationGracePeriodSeconds is 30.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[fawadkhaliq]

If preStop delay is set to a value greater than 30 secs, then terminationGracePeriodSeconds has to be adjusted accordingly

Looks good overall. In the description, can we please elaborate more on the 30 seconds with specifics. Why and where terminationGracePeriodSeconds needs to be changed. A sample pod spec would be great! Will be helpful for the users and anyone updating documentation