"non interruptible" forms (deferred posix signals)

[<don...@is...>]

 > Following a discussion on clisp-list with Don about
 > call-with-timeout behavior it looks good to introduce
 > "non-interruptible" forms.

I think you're missing something here.  The fact that a cleanup form
could be in an infinite loop suggests that we still want to be able
to interrupt it.  My first argument was that things like with-timeout,
at least by default, should complete all of the cleanup forms.
My second was that it seems important to be able to program a wide
variety of options.  This is why I wanted to separate the interruption
from the reaction.  I think adding a small amount of state would allow
the reaction code (which should be able to do to all that the debugger
now does), to do a wide variety of useful things, inc.
- interrupt and go into the debugger (despite any without-interrupts) 
- interrupt and execute only (and all) cleanup code out to the closest
  timeout, and return from the timeout
I suggest that those two are a bare minimum, even if you don't want to
support my more general scheme.