`letter_histogram`

     sub letter_histogram {
       my $strdex = (length $_[0])-1;
       $letter_hist{substr($_[0],$_,1)}++     for (0..$strdex);
     }

Perhaps we could get a speedup by avoiding the repeated array lookup on @_?

     sub letter_histogram {
       my $msg = shift;
       my $strdex = (length $msg)-1;
       $letter_hist{substr($msg,$_,1)}++     for (0..$strdex);
     }

Cost: shift plus an extra copy of the data

     Before                     After

     real    0m1.277s           real    0m1.236s
     user    0m1.250s           user    0m1.220s
     sys     0m0.020s           sys     0m0.010s

No significant difference
- Perhaps it really is .04 ms faster
- But who the heck cares?
Other things I tried:
- Use @letter_hist instead of %letter_hist
- Call letter_histogram once on entire mbox instead of on each message