add more disconnect/fork policies

[moshewe]

The current policy, as far as I understand from the source code, is to immediately fork |workers|/2 times (not including the last worker), immediately disconnect |workers|/2 workers, and disconnect the rest in one minute after which we fork again.

That's a simple yet somewhat brutish policy - forking up half your workforce is quite a load. I think a one-by-one disc/fork policy should also be enabled.