network start: Docker launcher discards container output on premature exit (same as #597 for native)

[lwshang]

Problem

#597 / #675 / #676 fix the native launcher so that when it exits prematurely (e.g. a fixed gateway port is already taken) the user sees the actual cause instead of just exited prematurely with status 101.

The Docker launcher path has the identical bug and is untouched. When network start runs the launcher in a container — always on Windows, and on unix with autocontainerize=true or a docker-mode network — it goes through spawn_docker_launcher, not spawn_network_launcher. On premature exit it returns ContainerExitedPrematurely with only the exit status, and never surfaces the container's output — even though it attaches stdout/stderr (attach_stdout/attach_stderr: Some(true)).

Suggested fix

In the premature-exit arm of the select! in spawn_docker_launcher (mirroring premature_exit_detail from the native path), fetch the container log tail via bollard's logs() API and fold it into ContainerExitedPrematurely. Note this is a different mechanism from the native fix (no local stderr.log file; the daemon owns the stdio), so it does not reuse premature_exit_detail.

Notes

• Deliberately scoped out of fix(network): surface launcher stderr on premature exit in background mode #676, which is unix/native-only and covered by an integration test that can't exercise the Docker daemon.
• Testing this path needs a working Docker daemon and is harder to make deterministic in CI.

Discovered while reviewing #675 and implementing #676.