job spends a long time doing things after a failed build

[Steve Langasek]

I was iterating on image builds to land a new feature for s390x in debian-cd, and I noticed that, even when all of the images that were to be built as part of the job had failed, cron.daily-live went on to waste time on some auxiliary actions that blocked the script from returning and therefore (effectively) blocked me from retrying.

An example is
https://ubuntu-archive-team.ubuntu.com/cd-build-logs/ubuntu-server/mantic/daily-live-20230928.3.log which shows:

[...]
make: *** [Makefile:896: /srv/cdimage.ubuntu.com/scratch/ubuntu-server/mantic/daily-live/tmp/mantic-s390x/bootable-stamp] Error 1
ERROR WHILE BUILDING OFFICIAL IMAGES !!
===== Publishing =====
Thu Sep 28 15:39:51 UTC 2023
No live-server image for s390x!
No images produced!
===== Purging old images =====
Thu Sep 28 15:39:51 UTC 2023
Purging ubuntu-server/daily-live images older than 1 day ...
===== Handling simplestreams =====
Thu Sep 28 15:39:51 UTC 2023
Refreshing simplestreams...
[...]

The 'refreshing simplestreams' blocked for several minutes on filesystem operations until I ^Ced the command. It should just never have been run at all when there were no new images produced.

Likewise, we shouldn't be running the "publishing" or "purging" stages when there were no new images - if this had been a regular daily job and images for all architectures had failed to build, we don't want to purge the previous good image when there's no new one to replace it!