fix(docker): restart container when backend exits

[samtruman]

Problem

When the backend exits intentionally for a settings restart, the web process remains alive under the shell entrypoint.

Root cause

The entrypoint waited for both background children, so PID 1 stayed alive and Docker could not apply its restart policy.

Fix

Detect the first child exit, handle zombie state, terminate the surviving child, and let the container exit so Docker can restart it.

Tests

• POSIX shell syntax check
• Isolated Docker lifecycle/restart test validated on the candidate image
• Existing suite: 258/258 pass