If `create_run_path` fails for any one realization (e.g. disk full) then experiment fails instead of one realization

When the disk is full and writing the run path for one realization fails, then the experiment fails instead of just one realization.

One solution would be to wrap the block of the loop

https://github.com/eivindjahren/ert/blob/2a72c438f347f662069a8d11fcafdb73eb68e6cf/src/ert/run_models/_create_run_path.py#L216

in a try-catch, and return which realizations one successfully created run path for then pass that to run_ensemble_evaluator in

https://github.com/eivindjahren/ert/blob/2a72c438f347f662069a8d11fcafdb73eb68e6cf/src/ert/run_models/run_model.py#L764-L794

Read src/ert/run_models/_create_run_path.py around line 216 and src/ert/run_models/run_model.py around lines 764-794. Trace how realization run paths are created and passed to run_ensemble_evaluator, then verify that a path-creation failure excludes only that realization while the remaining experiment continues successfully.

Written by the indexing model from the issue text.