Parallelize across genomic windows instead of full chromosomes

[jamorrison]

As of this writing, Tranquillyer parallelizes across whole chromosomes when deduplicating. Thread usage can be improved by parallelizing across equally sized bins within chromosomes. This will improve the consistency of run times across threads as well as improving thread utilization when more threads are created than chromosomes in the BAM file.

Note, this will require special care when handling reads near the boundaries of windows to avoid double counting reads and properly retrieving reads that fall within the window size around the read start position but fall in an adjacent bin (e.g., the start position of a nearby read is in the next bin, rather than the currently processed bin).