Summary

• Fix spawn server orphaning in cgroup-network when fatal() is called
• Prevents zombie processes when running as PID 1 in Docker

Root Cause

When cgroup-network calls fatal() after creating its spawn server (line 759), the spawn server process is orphaned and reparented to PID 1. When in Docker netdata runs as PID 1 (not by default), and when the orphaned spawn server exits, it becomes a zombie because netdata is not reaping non-child processes.

Solution

Register a fatal callback that properly destroys the spawn server before exit:

• Add cleanup_spawn_server_on_fatal() callback
• Register callback immediately after spawn server creation
• Set spawn_server = NULL after destruction in both cleanup paths (fatal callback + normal exit)

Safety Analysis

• The fatal() call (line 759) occurs before any fork() (line 825)
• No race conditions with parent/child cleanup
• Only one fatal() call exists in the entire program

Test Plan

• Verify compilation
• Test with invalid NETDATA_HOST_PREFIX to trigger fatal()
• Confirm no zombie processes remain in Docker

Fixes #20565