Waiting for ugpraded takes forever on double push

[svensp]

Waiting for the upgraded state takes forever if a version is pushed again while its current version is already in rancher.

Reason for this:
The created service definition is exactly the same as the one already on rancher. rancher-compose up --upgrade notices this and does nothing - the service is already running like this.
Thus the upgraded state rancherize is waiting for will never be reached because no upgrade takes place.
Note: This happens even if the image inside was changed because the service definition contains only the image registry+name+version, not its contents and thus still matches

Proposed Change:
Inside the InServicePushMode check the version which is currently online. If the new version is the same then throw an exception which warns about this problem.