Test Auto-Scaler functionality with single vm

[schmidt-scaled]

1. Provision with 4vcpu and 8GB of RAM
2. Measure the following settings AND actual allocation after "tuning" has been applied against the soft limit, but BEFORE any workloads: shared buffers, WAL buffers, work memory, maint. work memory, temp buffers, other (backend memory)
3. Apply now workload step by step in constant increments, but adding one session after another. each session runs constant load in a loop:
   create some table, insert fields, perform query, delete table.
4. Repeat 3. for multiple times, after each step, measure (2.) again AS WELL AS vcpu allocation
5. now do the opposite: remove one session at a time, measure (2.) again AS WELL AS vcpu allocation
6. now resize to 8vcpu and 16GB or RAM
7. check 2. again (vcpu should idle, but how much ram is consumed after resize particularly for shared buffers?)
8. now we prepare for and run a very complex query, which in theory can use a lot of ram, if running optimally
9. measure 2. during query and after query (15 seconds after, 2 minutes after)

If the vcpu or ram does not come back at any step, we need to interrupt the tests and analyse the issue.