The reason is simple.  When I check for docker-compose, the
execution can fail.  Then I'll check for the built-in
docker compose.  If set -e is active the first failure will result
the script being aborted with no chance to check for the second
option.