Post-harmonization QC reporting tool

[amc-corey-cox]

Build automated tooling to produce QC reports comparing harmonized output against reference data, based on the grading methodology developed in #196.

Inputs

• Harmonized output TSVs (from pipeline)
• Reference data summary statistics (e.g., TOPMed DCC aggregates or the raw data)
• Variable mapping configuration (transform-specs which harmonized variables to compare)

Outputs

• Machine-readable scores (JSON or YAML) — per-variable grades, sample counts, summary statistics, overall cohort grade summary
• Visual HTML report — human-readable summary with per-variable detail, distribution comparisons, and methodological notes

Scoring

Uses the grading rubric defined in #196 (A+ through D for both continuous and categorical variables).

Approach

• Start with one cohort (COPDGene — structurally simplest, established baseline)
• Reports should work as pipeline artifacts (generated alongside harmonized data)
• Summary statistics only — no participant-level data leaves the enclave

Context

Chris Siege developed and validated this comparison methodology manually across 9 cohorts. This issue is about automating it as reusable tooling. See #196 for the scoring rubric and framework details.

[amc-corey-cox]

Related: #312 (hv-lint cleanup/migration from HV repo) — pre-harmonization lint tooling that complements this post-harmonization QC work.