I mistakenly omitted this logic from the new algorithm.

Needed to initialize BEDCOV.zeroLength to false.

greatly simplified the mergeBed algorithm.  MUCH more elegant now.

1. A -f (overlapFraction) parameter that defines the maximum overlap
   that a randomized feature can have with an -excl feature.  That is, if a chosen locus
   has more than -f overlap with an -excl feature, a new locus is sought.

2. A new -incl option (thanks to Michael Hoffman and Davide Cittaro) that, defines
   intervals in which the randomized features should be placed.  This is used instead
   of placing the features randomly in the genome.  Note that a genome file is still
   required so that a randomized feature does not go beyond the end of a chromosome.

It now lives in the "filo" package.

Previously used time() which has 1 second precision.
Now use gettimeofday() which has millisecond precision.

complementBed now reports chromosomes that were not present in the BED file,
yet were in the genome file.  Thanks to Michael Hoffman for reporting this.

Fixed casting issue in windowBed.AddWindow() that cuased slop positions to run off the end of the chromosome.

This will eliminate the oft-reported compilation issue.