Wiping the database is:
- slow. It adds over a minute to the boot process of a docker compose up
- cumbersome. We start with a blank slate after every boot, which means
  issued badges, accounts, logins, etc all are gone.

Ideally the blank state is useful in itself, but it's lacking data and
setup currently. So after booting we need quite a few manual tasks each
time, to get relevant badges into users' backpack.

This feature was commented out in the commit:

  commit 2916276
  Author: Okke Harsta <[email protected]>
  Date:   Wed Aug 18 14:24:39 2021 +0200

      Temporary comment to fix deployment

  Δ apps/mainsite/management/commands/seed.py

But, as "temporary fixes" tend to do, it remained in, and now causes
work and delays when developing.

We also add a persistent volume to docker-compose so that the data for
the mysql database is stored outside of the container and can live
across container lifecycles.

All BadgeClasses in Medicine get alignment-items:
  - EQF
  - ECTS

This user is often used in demo's and in tests.

Currenly, failing seeds would be "eaten" by the try/except clauses, and
a calling script, e.g. the entrypoint.sh would think a seed was
performed successfully (no errors, exit code 0) and continue.
But that would start a server that does not have required data.

This change ensures errors are propagated to the caller of the script
and exit code is nonzero if a seed fails.

WIP, because we need to consider potential issues for production and
deployments that this might cause. If any.

Also present in 0ab582f, but that hasn't been merged yet.

…ment

Instead, pick the first, and if none exist, crash.

We fix this by explicitly setting an ID, and by allowing multiple
accounts with the demo emailaddress to exist. We then pick the first.