Introduce methods for standardized callset evaluation

[kjaisingh]

Description

This PR is intended to provide a series of changes to methods used to do callset evaluations, including the following components:

• Introduces a new WDL for callset evaluations.
• Add stratification options to SVConcordance.wdl, which are now used in the callset evaluations.
• Allow Vapor to be run on CRAM files stored in requester-pays buckets.
• Provide option to modify the SV size limit for Vapor in the long-read comparisons.

Testing

• The following two workspaces (DRAGEN-SV-Comparison and DRAGEN-SV-Evaluation) highlight successful use cases of each of the WDLs and scripts.
• Validated all WDLs with womtool.

Pre-Merge Changes Required

• Create a new callset evaluation WDL.
  • Use the modified MakeGqRecalibratorTrainingSetFromPacBio as a base.
  • Rename MakeGqRecalibratorTrainingSetFromPacBio to use a prefix of AllOfUs and deprecate it into the internal repository.
  • Modify input parameters to be agnostic to the type of comparison needed: raw calls, joint callset, CNV sites or CNV calls.
  • Add input parameters to optionally do Vapor preprocessing, if Vapor is required.
  • Add input parameters to optionally standardize input VCFs.
  • Remove differentiation between loose and strict modes.
  • Integrate pre/post-processing functionality from all the various new scripts & WDLs into this.
• Include a Python notebook for the analysis of results from the callset evaluation WDL, providing user-level parameters that are intended to govern how to process the data output by the evaluation WDL.
  • Modify input parameters to be agnostic to the type of comparison needed: raw calls, joint callset, CNV sites or CNV calls.
  • Use the pacbio_support_summary_table for Vapor results instead of the vapor_output_json.
  • Generalize the stratification field definition.
  • Include optional sections that can be run depending on what comparison mode we are in - e.g. short-read caller concordance, caller & depth only coverage, subsetting results to samples etc. Look through archived slides for all the different comparisons done for ideas for these.
• Add documentation that outlines how to use the long-read evaluations.
• Decide on the future of the DRAGEN-CNV standardizer (std_dragen_cnv.py) → if it is being included, add it to the standardizer options and documentation.
• Remove automated sync of the MakeGqRecalibratorTrainingSetFromPacBio WDL to Dockstore.