Consistent handling of existing ensembles and experiments in "Ensemble experiment"

[oyvindeide]

GIVEN case exists with parameters, no responses, press play on Ensemble Experiment
Option 1) Overwrite existing case with new sampling of parameters
Option 2) Evaluate the existing prior

There are currently two workflows being executed with Ensemble experiment, one which is a design matrix workflow, expects resampling of parameters, and does not really care about ert internal storage (we should change that fact, but that is another issue).

The second is a workflow evaluating either a sampled prior or the result from a manual update step. These are very distinct, and we should make sure work as expected, with no unexpected overwriting of parameters.

[oyvindeide]

Closing as duplicate of: #5686