Simplify everest server launching

[frode-aarstad]

Everest server currently starts an entire ERT instance only to run itself as a forward model. We do need to submit the everserver to the queue, but how we do it is up to us. One way of doing this could be to do a BaseRunModel, but BaseRunModel might make more sense for the actual batch simulation / optimization part.

(Original text: Create a BaseRunModel for the everest server.)