Job that exit hang (part of the) pool

[haf]

Let's say you have a function like

let rec f i =
  job {
    if i > 4 then () else
    do! timeOut 300
    return! f (i+1)
  }

And you >>=. (f 0) that, you'll indefinitely block the worker thread from taking any further work.

Similarly for exceptions, it would seem. So I suggest we track that all worker threads can be used, or even crash the app with a nice stacktrace if an unhandled exception or recursion like above ever reaches the library. Can we heuristically track "all worker threads are either executing jobs or waiting on a semaphore/monitor/WaitOne?

We should refactor towards this:

1. Add the ability to hook an event on fatal exceptions (anything uncaught); this should be used for logging fatal/critical errors in production
2. Handle "lost" threads

[TheAngryByrd]

I'd definitely be interested in finding out where worker threads get blocked.

[haf]

Added some comments; I think this would be a nice first PR to Hopac for someone; trigger an event upon a job crashing.