Document the impact of a buffer on a small queue of jobs when a job takes a long time

[ZimbiX]

From this issue:

If you only have a very small number of jobs currently in the queue, try decreasing the buffer size. We saw that kind of issue early on in us using Que - we had one long job which which was blocking idle worker processes from picking up the couple of quick jobs which were in the queue behind it. iirc, because a worker process by default allocates several jobs to itself (the buffer) as a performance optimisation. We should probably document this behaviour. And remove the minimum buffer size option, as I don't think that one's actually used anymore.