[plt-scheme] mzc: native code path + plt/collects/mzlib

   >  But wouldn't you think that row reducing matrices and
   > quicksorting long lists would profit from native compilation?

   Isn't merge sorting faster on large lists?

Maybe so, Jens!  And maybe my claim was false, since I do merge
sorting too, and since that's my own code, it's compiled.  So maybe I
don't quicksort compiled as well, and as Matthew F said, quicksort
might not benefit from compilation anyway.

Anyway, I was more worried about running background jobs than speed.
With Dave's `nohup' fix, I was able to run a job last night (on a
public machine after I logged out), and got this output (using
current-seconds to time the `s' calculations):

(t = 56) 
(s = 0) 0.0 minutes
(s = 1) 0.0 minutes
(s = 2) 0.0 minutes
(s = 3) 0.0 minutes
(s = 4) 0.0 minutes
(s = 5) 0.0 minutes
(s = 6) 0.0 minutes
(s = 7) 0.3 minutes
(s = 8) 0.1 minutes
(s = 9) 86.3 minutes
(s = 10) 18.8 minutes
(s = 11) 3.4 minutes
(s = 12) 6.1 minutes
(s = 13) 5.3 minutes
(s = 14) 5.6 minutes
(s = 15) 9.0 minutes
(s = 16) 0.6 minutes
(s = 17) 23.0 minutes
(s = 18) 1.8 minutes
(s = 19) 0.7 minutes
(s = 20) 0.0 minutes
(s = 21) 0.0 minutes


I can live with that.  It's more of a slippery slope: a background job
means mzscheme instead of drscheme anyway, so why not compile?  But I
oughta cvs process.ss to be sure.  Thanks, everybody!