Currently, E2E tests are highly dependent on the k3d cluster setup. The tests are creating resources based on the count of nodes and their memory size.
It makes life harder, to reuse those test in different kind of environments (Kind, Cloud based, Qwok, etc)
This refactor should defien the interface between the cluster setup and the tests, and based on this configuration interface tests should adjust the parameters or skip gracefully if there are not enough resources