Capability to cancel dask jobs started with fire_and_forget

Ran across a scenario where a job was improperly submitted using "fire_and_forget" to my dask cluster that was too small/wrong level of parallelism and was going to take a long time to finish. The only way I could find (out of the box) to cancel the job was to blow away the whole cluster and restart it.

It would be nice if Coordination Primitives as described in this (closed) issue:
https://github.com/dask/dask/issues/1183

Could be combined with an approach like described here (explicit stopping condition):
https://stackoverflow.com/questions/49203128/how-do-i-stop-a-running-task-in-dask

To give out of the box ways to have Client apis like:

```
Client.get_jobs
>> Queue[Future]

Client.cancel_job(future.id)
>> Cancelled
```