child_cleanup: reap as many child processes as possible
Under load it's possible multiple child processes have been killed before
we start processing the SIGCHILD signal, leaving zombie processes behind
everytime we miss a process.
Reap as many processes as possible instead of assuming one handler
call = one process like we currently did.
Can be reproduced by running the following commands in parallel:
- tcpdump -i lo -w /tmp/test -C 1 -z /usr/bin/true
- iperf3 -s
- iperf3 -c localhost
(cherry picked from commit
b5f0be2196d897c76144a5b90b168836df54d6fa)