Compose up cycles all services #1286
Description
Right now, defang up will cause all (unmanaged) services to be cycled. This is mostly because all container names will get updated to reflect the new deployment ID (Etag), so their logging can be captured with defang tail --deployment 123. It's very annoying to have all services cycle, because of the inter-service dependencies. They all cycle at the same time, so there's a lot of races to make sure one service can connect to another, causing further crashes and restarts, until the cluster stabilizes.
Current AWS CD code adds deployment ID/etag to each container name, so the ECS lifecycle events contain the etag. There might be a way to link the ECS deployment events to the ECS task events.
If a service depends on config values and those values were changed (version numbers), then we should cycle the service to ensure latest config is applied.