raise severity of non-empty docroot test

previously, if the docroot was non-empty, the entrypoint script would wait for 10 seconds and then continue. now, it will simply fail. *unless* the `SKIP_CHECK_EMPTY_DOCROOT` envvar is set to some non-empty value. this is a somewhat "breaking change", as people with half-popuplated docroots (e.g. having an local skin; like me) will now experience failures, unless they manually add the `SKIP_CHECK_EMPTY_DOCROOT` envvar to their setup. - Closes: https://github.com/roundcube/roundcubemail-docker/issues/384