v2.0.0

This is the 2.0 release. It includes numerous features and bug fixes compared to 1.0, including:

• It supports soft clipping of reads based on both the alignment and base-call quality scores, using Hamming distance to determine when to do this.
• It supports liftover of alignments from ALT contigs to the corresponding regions of primary contigs
• It has the ability to quickly find large indels in the middle of reads by leveraging the pattern of index hits to detect where this might be happening and looking more carefully at those possible alignments
• It will by default report reads with two equally good possible alignments at MAPQ 0 rather than 3, because variant callers expect this regardless of the definition of MAPQ in the spec
• It has a mode that allows it to work well with the DRAGEN variant caller (this is just enabling a set of options that work well with DRAGEN but less well with Haplotype Caller)
• Some different default parameter settings
• 10% faster performance on some high-quality read sets that we measured.
• Many fixes for correctly calling short indels, sometimes by employing base call quality scores
• removing many memory leaks
• making the output BAM file have its contigs in the same order as those in the FASTA used to generate the index (though not required by the spec it is needed by the DRAGEN variant caller)
• fixing some problems in CIGAR string generation
• eliminating some problems writing very large BAM files on Linux
• not binding threads to processors when affinity is already set at process start time (Linux only)
• improvements to error messages