closestBed does some extra filtering of the hits returned by BedFile::allHits, but
it was previously assumed that each hit from allHits would be counted by closestBed.

This fix adds a check for the -t to make sure there are some hits in closestB and
updates numOverlaps when hits are filtered.

Maybe it would be better to add orientation-specific search to BedFile::allHits
instead of in closestBed?

 * requires -D and reuses its logic to determine orientation

Improvements to closestBed.  Thanks to suggestions by Brent Pedersen, Ryan Layer, Assaf Gordon, Dan Webster.

1. All overlapping features are reported by closestBed -t all.
2. The "-io" option will prevent overlapping features from being reported.

	1. Reset BED.strand to be string instead of char in the interest of generality.
	2. Fixed bug in genomeCoverageBed where NULL entries where genome summaries were err. reported with -bga.
	3. Fixed bug in closestBed where NULL entries were err. reported when BED > 6.
	4. Standardized the Open, Get, Close idiom per improvements from Gordon Assaf.
	5. Add "split" logic to bamToBed.

	1. Revised usage to 80 chars
	2. GPL headers
	3. Added new stdin logic

	Fixed a bug in closestBed that caused all nearby (not just closest) features to be returned.

=== Notable changes in this release ===
1.  coverageBed will optionally only count features in BED file A (e.g. sequencing reads) that overlap with
	the intervals/windows in BED file B.  This has been requested several times recently and facilitates CHiP-Seq and
	RNA-Seq experiments.

2.  intersectBed can now require a minimum __reciprocal__ overlap between intervals in BED A and BED B.  For example,
	previously, if one used -f 0.90, it required that a feature in B overlap 90% of the feature in A for the "hit"
	to be reported.  If one adds the -r (reciprocal) option, the hit must also cover 90% of the feature in B.  This helps
	to exclude overlaps between say small features in A and large features in B:

	A ==========
	B  **********************************************************

	-f 0.50 (Reported), whereas -f 0.50 -r (Not reported)

3.  The score field has been changed to be a string.  While this deviates from the UCSC definition, it allows one to track
	much more meaningful information about a feature/interval.  For example, score could now be:

	7.31E-05  (a p-value)
	0.334577  (mean enrichment)
	2:2.2:40:2 (several values encoded in a string)

4.  closestBed now, by default, reports __all__ intervals in B that overlap equally with an interval in A.  Previously, it
	merely reported the first such feature that appeared in B.  Here's a cartoon explaining the difference.

	**Prior behavior**

	A	 ==============
	B.1        				++++++++++++++
	B.2       				++++++++++++++
	B.3               				+++++++++

	-----------------------------------------
	Result = B.1 			++++++++++++++

	**Current behavior**

	A	 ==============
	B.1        				++++++++++++++
	B.2       				++++++++++++++
	B.3               				+++++++++

	-----------------------------------------
	Result = B.1 			++++++++++++++
			 B.2 			++++++++++++++

	Using the -t option, one can also choose to report either the first or the last entry in B in the event of a tie.

5.  Several other minor changes to the algorithms have been made to increase speed a bit.

1. Fixed yet another bug in the parsing of "track" or "browser" lines.  Sigh...
2. Change the "score" column (i.e. column 5) to b stored as a string.  While this deviates
   from the UCSC convention, it allows significantly more information to be packed into the column.

	1. Added strandedness to relevant programs (e.g. intersectBed, mergeBed, etc.)
	2. Moved reportBed and reportBedRange to bedFile.cpp to clean up code
	3. Sped up file parsing.  Added lineFileUtilities.cpp
	4. Added peIntersectBed