Ticket 374: Treat unknown exit argument as #f

Currently, it's implementation-dependent what an argument to exit other than #f means. Ticket #373, if it passes, will also define #t.

This ticket doesn't change that, but just adds language to the effect that if the implementation doesn't know what to do with the argument (i.e. a string on Windows or Posix, or a procedure in any existing OS), it should treat it as #f rather than possibly failing to exit.

The reason to choose #f is that there is in most OSes only one way to express success and many to express failure. So if an argument is uninterpretable, we can at least interpret it as failure.