[Shootout-list] A few unrelated points

[Stephen Weeks]

I checked in the MLton sumcol improvement suggested by Ray Racine.
Thanks!

I support the idea of changing the sumcol input to include negatives
so that bignums are not required.  In the interim, while it is still
being decided, I think the test should be disabled.

The ghc regexmatch benchmark doesn't do the same computation as other
implementations and never has.  It should be disabled until it does.
It is misleading and confusing to (the many) people who do not know
about Haskell and laziness how ghc can be so fast on this benchmark.
Furthermore, it is easy to modify the other language implementations
to use laziness as well -- there is no reason to give Haskell a free
pass just because its laziness is in the language rather than a
library.  If ghc is not disabled, then I think it is reasonable for
other language implementations to change their code to avoid
recomputing the match as well.

The bench/reversefile directory now includes a 101M tmp directory.
This seems like it shouldn't be included in the CVS.