Decide whether memory-pressure regrading becomes the default for servers #140

The memory-pressure regrading service (github_runner_cluster_regrade) defaults to false. A fleet of three Docker-on-Mac servers now runs it, so it can be judged before the default changes. Decide whether to make it the default for a server, once the fleet has seen at least one real step down and step up: that the Lease, drain, etcd-only stop and automatic step back up worked end to end under real pressure, that no step happened without pressure, and that a server whose kernel lacks /proc/pressure/memory or whose group has a single server is handled by the role's own refusal or stop, so a default of true cannot break a deploy that worked.

Deferred until the trial has produced that evidence. If the default changes, every consumer's docs and the role's assertions for agents and single-server groups need to read cleanly under it.