Release 4.0.0

• Updated default versions for several tools to the newest version. Pin the
  volatile python container to a specific hash digest.
• Changes in Cromwell 48 made it impossible to use the wide array of inputs
  in our documentation (such as
  Germline.sampleWorkflow.QC.Cutadapt.minimumLength). Fixes have been made
  upstream and in the pipeline. From Cromwell 52 onwards these options are
  available again.
• WDL files and import zip packages are now offered with every release to
  make downloading and running much easier.
• The output directory was restructured to contain less nested folders.
• Performance has improved as a result of extensive benchmarking and profiling.
  Details can be found here.
• Added a bwaThreads option to the input. This sets the number of threads used
  by the BWA aligner, and is a major influencer of wall clock time.
• Use scatter-regions from the chunked-scatter python package as scattering
  tool. The old biopet-scatterregions package did not support scattersizes
  larger than 2 billion, making it impossible to have the whole human genome in
  one scatter.
• Use sambamba markdup for marking duplicates with 2 threads which reduces
  wall clock time.
• Updated default BWA containers. Both bwa and bwakit now contain 0.7.17
  and include samtools 1.10 for fast sorting.
• Added bcftools stats to the pipeline for variant
  statistics. These stats are reported in the MultiQC report.
• Tasks were updated to contain the time_minutes runtime attribute and
  associated timeMinutes input, describing the maximum time the task will
  take to run.
• Document the use of cromwell's final_workflow_outputs_dir feature which
  makes the germline and somatic pipelines usable on all of Cromwell's
  supported backends. Users are encouraged to use this feature. outputDir references are removed from the documentation.
• Make the MultiQC task suitable for use with a final_workflow_outputs_dir
  so it can be used on all of Cromwell's supported backends.
• Restructure the pipeline so variant calling jobs are scheduled more
  efficiently.