Try to run and build tales on the same node

[craig-willis]

In a multi-node configuration, if I restart or modify and rebuild my tale chances are the task will end up on a different node. This will require pulling the image or rebuilding without the benefit of the cache. If the node is still available and we know where the image was last run or built, we should send it there. This would mean storing something like lastKnownInstanceNodeId or lastKnownBuildNodeId and using it as well as handling when the node isn't there.