Use generated datasets in all nightly performance test pipelines

Context:
We currently have two "large test" pipelines that are using a rather outdated large dataset (1Mio ProcessInstances) originating from a 24/7 running postgres instance. We should rather migrate them to use the data dumps generated with dumps created by the `Generate Camunda BPM test datasets` job.

These jobs are:
Upgrade performance test on large static dataset
Query performance test on large static dataset

AT:

• the mentioned pipelines use the large data dumps rather than the postgres instance
• the not required postgres instance is eventually shutdown by the INFRA team