improve bw/entrypoint.sh behaviour

[ppenguin]

Remove infinite restart-loop with incorrect condition; make container exit with rc 1 if main nginx exits uncleanly.

I'm not 100% sure whether this leads to the desired behaviour yet (should be tested).

Before: if the nginx process crashes, wait $pid stops blocking, but the crashed proces doesn't cleanup the pid file so the while loop does nothing and causes 100% load.

After: if the main nginx proces exits without cleaning up the pid file, it is considered an unclean/unexpected exit and causes the stale pid file to be cleaned up by the entrypoint and the container exiting with status 1.

However: if the container orchestrator is configured to restart containers on failure, this would cause a new bw instance to be started, so far so good. BUT: will this newly started container request the needed configs from the scheduler upon startup? Or is configuring "push only" (scheduler -> bw)? In the latter case we still have a problem.