Support multi-node task-queue

[timbru]

Depends on #1101.

When multi-node setups with database storage (postgresql) are supported, then we also need to support multiple nodes using a shared queue.

We may want to insist that each node has a unique name (if postgresql is used as a storage option) and track which node is running a certain task.

Then we can also use this info on a restart. I.e. all running tasks for the restarted node can be assumed to have failed and can be restarted.

[timbru]

Closing this for now. The task queue relies on storage transactions, at least in #1152. This should work cross node if a database is introduced. If there is a desire to do something else then it should probably be discussed elsewhere, and then issues can be created as needed.