The new grace period defines how long Traefik should continue accepting
requests prior to shutting down all connection listeners. It preceeds
the existing (HTTP shutdown) grace period and allows for a full-circle
graceful termination by expecting downstream load-balancers to take
Traefik out of rotation in time, without the need for rather complex and
external request draining procedures.

Immediate beneficiaries are providers such as Kubernetes and Marathon
which expect in-cluster applications to exhibit the described
termination behavior.

Former name stems from a previous (never published) naming schema.

Doesn't really make sense since we shut down the server for SIGINT
gracefully as well.