create reasonable set of test data

[fwilhe]

currently the image uses the 'fulldata' variant which contains a full dump of all the ingestion data.

this is too large to be handy and also too large to be committed to git in any reasonable way.

creating a useful slimmed-down data set is not trivial because of the dependencies between the table. some automated way to only keep the relevant data for unit-test purposes would be nice.