Make job_runner.py internalise results directly

One option is making internalising results a separate job, which has the added benefit of visualising it in the gui as well. This means ert should send the location where the data should be internalised as part of jobs.json. Ideally we would like to send this over the network, but currently that would not solve any of the problems we are seing.

Start by reading job_runner.py and tracing how jobs.json is produced and consumed. Clarify whether internalising results should be a separate GUI-visible job and how the data location should be represented. Done means ERT supplies that location through jobs.json and job_runner.py internalises results directly; network transfer is noted as a future concern.

Written by the indexing model from the issue text.