runtime.update can block all checkins and state processing while waiting for shutdown

[leehinman]

update can block because it waits for waitForStopped to finish. waitForStopped can take up to 15 seconds. Because this function is called by the Coordinator, the entire Coordinator can block for that time.

This block causes problems because the runtime Manager update channel needs to write to coordinator.watchRuntimeComponets. If it cannot do that it will block. This in turn blocks runtime.stateChange which prevents runtime.runLoop from processing checkins and state changes, leading to stalled communication and "missed checkin" errors for components that are still sending checkins.

This can lead to the bugs seen in #3617 and #3654

[elasticmachine]

Pinging @elastic/elastic-agent (Team:Elastic-Agent)