futures and error dodginess

[Kevin Ryde]

Running

    (future-ref (future (error "x")))

sometimes gives

    Backtrace:
    In unknown file:
       ?: 0* [primitive-load "/home/gg/test/future.scm"]
    In /home/gg/test/future.scm:
       2: 1* [future-ref ...
       2: 2*  (future (error "x"))
    In unknown file:
       ?: 3*  [#f]

    /home/gg/test/future.scm:2:13: In procedure apply in expression (future 
(error "x")):
    /home/gg/test/future.scm:2:13: Wrong type argument in position 1: #f

It seems to be a race, sometimes the error is from future-ref about a
failed future (which is good), or sometimes just a segv.

I suspect %threads-handler is its dummy #f when future_handler tries
to call it.  Maybe that definition should be in boot-9 instead of the
threads module.

Adding that module

    (use-modules (ice-9 threads))
    (future-ref (future (error "x")))

then gives a printout

    In thread:
    ERROR: x

but hangs.