Separate ecl and flow simulator logic

[DanSava]

Issue

When a config file contains the FLOW forward model job, running of the simulator is handled by EclRun and it's method runEclipse.

We would like to have Flow and Eclipse simulator handling code separate.

eclrun is currently responsible for running both FLOW and Eclipse, and is dependent on environment variables that may or may not be overwritten by ert-configurations. This could be split up quite a bit. Suggestion: Without ert-configurations means it is outside the EQN cluster, with ert-configurations means it is on the cluster. Further we can split into FLOW and eclipse.

	On cluster (has ert-configurations)	Outside cluster
Eclipse	ecl_run	ecl_run_custom
FLOW	flow_run	flow_run_custom

It is somewhat clear that this logic needs to be split up, but some decisions in structuring this depend on some questions that are up for discussion:
Q: What should be the responsibility/purpose of eclrun, and what is not the responsibility/purpose of eclrun?
Q: Should eclrun be for only EQN / cluster users?
Q: Where should the logic to check for ecl version exist? (If we are to put it inside our ecl forward model, it needs to be within ert). Should it be only for on-cluster, or outside cluster only?
Q: Should all our FLOW/EclipseX00 forward models go through what is now ECLRUN?

[sondreso]

There are ongoing discussions on how we integrate with FLOW, I suggest we put this issue to the backlog until that is resolved as some of the answers to the questions above depends on the outcome of this discussion 🙂

[berland]

Superseded by #8925