[Termtools] timeout while writing output

[Johannes Waldmann]

I noticed a few cases where matchbox was apparently
killed by timeout while writing output, like here:
http://nfa.imn.htwk-leipzig.de/termcomp-2015/pairs/121496945
http://nfa.imn.htwk-leipzig.de/termcomp-2015/pairs/121442140
It's still counted as YES by the post-processor.

NB: Why does it take several seconds to output a proof trace?
I think one thread found the proof and started to print,
but the (Haskell) run-time system interleaved
the printing with other threads that were still running
and somewhat hard to kill, like FFI call to SAT/SMT solver,
or tight (nonallocating) Haskell loop.
I am already using +RTS -C because of that.

- J.