Add GATK4 components

[lazappi]

Description

Add components for functions in the Genome Analysis Toolkit (GATK4):

• gatk4/gatk4_createsequencedictionary: Create a sequence dictionary (.dict) for a reference FASTA file
• gatk4/gatk4_markduplicates: Locate and tag duplicate reads in a BAM file
• gatk4/gatk4_baserecalibrator: Generate a base quality score recalibration (BQSR) table from known variant sites
• gatk4/gatk4_applybqsr: Apply a base quality score recalibration table to a BAM file
• gatk4/gatk4_haplotypecaller: Call germline SNVs and indels from a BAM file via local re-assembly of haplotypes
• gatk4/gatk4_combinegvcfs: Combine per-sample GVCFs into a single multi-sample GVCF
• gatk4/gatk4_genomicsdbimport: Import per-sample GVCFs into a GenomicsDB workspace for scalable joint genotyping
• gatk4/gatk4_genotypegvcfs: Perform joint genotyping on GVCFs, combined GVCFs, or a GenomicsDB workspace
• gatk4/gatk4_variantfiltration: Filter variant calls based on INFO and/or FORMAT annotations
• gatk4/gatk4_selectvariants: Select a subset of variants from a VCF based on various criteria

Checklist before requesting a review

• I have performed a self-review of my code

• Conforms to the Contributing guidelines

• Proposed changes are described in the CHANGELOG.md

• I have tested my code with viash ns test --parallel -q <name or namespace>

• Check the correct box. Does this PR contain:

  • Breaking changes
  • New functionality
  • Major changes
  • Minor changes
  • Documentation
  • Bug fixes

[mdgrv]

argument changed names:
The real mutually-exclusive partner (per the GATK help dump) is this component's --quantize_quals.

[mdgrv]

rm whitespace after |

[mdgrv]

Every VCF (VCF, GVCF) matches the "^##fileformat=VCF" header line. Maybe use the "##GVCFBlock header lines to check?

[mdgrv]

The test cases use GVCF, the gendb:// is not included in the genotypegvfs tests. Did a real run with gendb:// and all runs fine, there is just no test to catch future breaks.

[mdgrv]

Maybe add a check to see if with --remove_duplicates the BAM output is lower than the output in test 1 (default, w/o removal of duplicates)? We need to have less read than default run in case there are duplicate reads present.

[mdgrv]

I reviewed all 10 components statically, ran the full viash ns test -q gatk4 suite, and exercised the entire pipeline end-to-end on real data (SEQC2 HCC1395 chr21, ~30x).

Checked:

• Static (all 10): correct viash 0.9.x conventions, DRY utils/ layer, meta_* (never par_*), JVM -Xmx from meta memory, multiple:true via split_multiple_to_flags, helper-based staging, genuine 4.6.2.0 help dumps, CHANGELOG cites PR #226, ASCII-clean.
• Tests: all 10 build + test.sh pass; ASCII build gate green for all 10.
• Real-data end-to-end: the full chain (markduplicates → baserecalibrator on real bgzipped dbSNP138 → applybqsr → haplotypecaller GVCF → combinegvcfs/genomicsdbimport → genotypegvcfs from both gendb:// and combined GVCF → variantfiltration → selectvariants), every step succeeded. Highlights: gendb vs combined give identical genotypes; paired filter expr/name map correctly; SNP selection works.

Suggestions (up to you, see inline comments):

• Test-coverage gaps — haplotypecaller GVCF test only checks the generic VCF header (assert <NON_REF>/##GVCFBlock); genotypegvcfs gendb:// branch untested; markduplicates --remove_duplicates doesn't assert dropped reads.
• Doc edits — applybqsr --round_down_quantized references a nonexistent --quantization_levels (should be --quantize_quals); variantfiltration trailing whitespace after |.