Update CombineBatches workflow

[mwalker174]

Replaces most methods in the CombineBatches workflow with a greatly simplified set of tasks that utilize GATK SVCluster and the new GroupedSVCluster tool (see PR). SVCluster replaces most of the current functionality including VCF joining and clustering, while GroupedSVCluster introduces refined clustering (a.k.a. "reclustering") that has become a best practice for larger call sets.

Clustering refinement is critical for consolidating redundant variants in repetitive sequence contexts such as simple repeats and segmental duplications. This also addresses an issue with duplicate insertions that share coordinates but have slightly different split read signatures (i.e. different END positions).

In addition, this PR makes some minor improvements to VCF formatting and parsing:

• In GenotypeBatch, CNVs are now formatted the same way as in CleanVcf, i.e. no genotypes and <CNV> ALT allele and SVTYPE=CNV, rather than using alt alleles <CN0>,<CN1>,… and having SVTYPE=DUP. In case a user needs to run on a VCF with the old format, there is a legacy_vcfs flag in CombineBatches that will update to the new format prior to processing.
• Within CombineBatches, records are annotated with HIGH_SR_BACKGROUND and BOTHSIDE_PASS INFO field flags rather than passing around separate lists, which is cumbersome.
• Minor improvements to some downstream scripts to use .get() for accessing FORMAT fields rather than brackets. This was required in some cases because GATK omits a FORMAT field if it is null for all samples in a given record. Pysam then throws an error since the requested key does not exist, whereas .get() returns None.
• The VCF is converted back to the "old" format at the end of CombineBatches to minimize risk of bugs in downstream workflows.
• Minor change to the breakpoint overlap filter: variants are prioritized on BOTHSIDE_PASS status (binary) rather than fraction of supporting batches.