[imikejackson]

Summary

This PR holds all V&V work for the filters that have a V&V record: the V&V policy and templates, the V&V records, and the code fixes for defects found in V&V'd filters. It is a draft. Do not review it yet.

The work stays on one branch until the V&V policy and the report template are stable. This prevents a series of merges that later template changes would make out of date.

V&V policy and templates

• The V&V reports follow docs/vv_templates/report_template.md exactly.
• A report is written for a user of develop or of a DREAM3D-NX release. It names DREAM3D-NX release versions, not branches, commits, or PRs.
• Deviation entries use only the template fields and one optional Evidence paragraph.
• The sign-off form is <Name>, YYYY-MM-DD; second engineer: <Name>, YYYY-MM-DD. The second engineer is the reviewer who approved the PR with the V&V work. When no approval exists, the sign-off says "second engineer: pending" and the Status is READY FOR REVIEW.
• The defect-hunt policy and a part for SIMPLNX defects are added to the templates.
• docs/vv_templates/vv_recertification.md says when V&V work happens after a filter has a V&V record. The trigger rule (does the change alter output for valid input?) is proposed for team review. The single-branch rule and the legacy comparison package rules are agreed.

V&V records

• All 44 V&V records are rewritten to the template.
• The DBSCAN record, deviations, and provenance updates move here from BUG: Enforce component shapes and types that filters read #1766.
• New legacy comparisons (DREAM.3D v6.5.171, a local proof version, and DREAM3D-NX) are recorded for Compute Feature Neighbor C-Axis Misalignments, Compute Feature Neighbors, Compute Feature Centroids, Compute Feature Neighborhoods, and Regularize Z Spacing. The comparison packages are in the V&V working archive.

Fixes in V&V'd filters

Each filter fix is a separate commit with tests that fail on the old code. The fixes are separate from the documentation commits, so they can move to their own PR if necessary.

• Compute Average C-Axis Orientations: bound Feature Ids at execute time (-76405) and accumulate in double.
• Compute Feature Reference C-Axis Misalignments: bound Feature Ids (-9806), use the same cell rule for the mean and the standard deviation, and accumulate in double.
• Compute Feature Centroids from Triangle Geometry: reject empty or mismatched Face Labels (-12903, -12902) and accumulate in double. The exemplar tolerance changes from 0.0001 to 0.05 because the archive holds float32-accumulated centroids; a float64 reference from the archive triangles differs from them by at most 0.0415.
• Isolate Largest Feature (Identify Sample): fill holes on 2D images.

Status

• Depends on BUG: Enforce component shapes and types that filters read #1766: the DBSCAN record describes tests that BUG: Enforce component shapes and types that filters read #1766 adds. This PR must merge after BUG: Enforce component shapes and types that filters read #1766.
• More fixes for V&V'd filters will be added, one filter at a time.
• The V&V records of the four fixed filters will be updated after their legacy comparisons run on the fixed code.

Test Plan

• Full test suite passes serially (in-core Release build, PY:: tests excluded).
• Each changed V&V report matches the report template.
• Every deviation ID in a report has an entry in its deviations file.
• Out-of-core test suite passes.