Svtyper

- Set --max_reads to unlimited by default, as a max of 1000 was conflicting with some of our pipeline steps

- New default behavior clears existing QUAL scores. (Use the -q flag to iterate existing QUAL field from input VCF)
- Limit number of reads to 1000 at each variant site (--max_reads), to prevent slow processing time in high-depth, repetitive regions (thanks to brentp)
- add argument (-T) for optional reference FASTA when reading CRAM files. (This prevents automatically downloading of the reference cache)
- Throw error if SVTYPE missing from VCF file

- Fixes a bug in building insert size distributions from CRAM files

This release fixes a bug in v0.1.0 where support for non-destructive events (DUP and INV) might not be appropriately counted. It is also slightly more specific about including soft-clipped read support which may reduce support relative to v0.1.0 in some cases.

- No longer requires MC (mate CIGAR) or MQ (mate mapping quality) tags from Samblaster
- No longer require splitters.bam file from Samblaster. All split-reads are filtered internally using the BWA-MEM SA tag
- Can read (and write) library information as a JSON file using the `-l` flag
- Can output a sparse BAM file of the reads used for genotyping with the `-w` flag (for diagnostic purposes).
- Much more memory efficient
- ~50% faster than v0.0.4
- Can read CRAM files as input
- Now assesses clipped reads (in addition to split reads) to call breakpoint genotypes

Many thanks to ernfrid for help with development, testing, and benchmarking for this release

Bug fix for https://github.com/hall-lab/svtyper/releases/tag/v0.0.3 for a serious bug introduced in 9dd73d2 that caused SVTyper to ignore paired-end evidence. Patched with a2a9381 (which is included in this release