Expanded processing functionality

We need better capabilities of hooking into the processing of deployments. When running two separate sites for content staging and production a lot of things might need to happen before or after deployments:

Also, we should separate the deployment to two stages:

1. Deploy all entities in the plan to the endpoint only creating new unpublished revisions (if supported for that entity type)
2. If step 1 is successful: Publish all revisions that were deployed (if supported for that entity type)

Reason: If something fails in the middle of a deployment, we limit the damage of changing published entities in a failed deployment. We could possibly add a third step that deletes all revisions that were deployed in a failed deployment. But we'll leave it for now, since it could be good to keep that record for consistency and logging purposes.

All this requires some changes to the deployment process. This issue will be a meta issue for this.