ENH: Add Use Feature Ids option to ComputeKernelAvgMisorientations

[imikejackson]

Summary

Adds a use_feature_ids boolean parameter (default: true) to the Compute Kernel Average Misorientations filter.

• Checked (default): unchanged per-grain KAM — kernel neighbors contribute only when they share the focal cell's Feature Id. Verified behavior-identical to the previous implementation.
• Unchecked: per-voxel KAM — the average may cross grain boundaries; a neighbor contributes when it is in-bounds, has Feature Id > 0, and matches the focal cell's Phase. The focal-cell validity gate (Feature Id > 0 and Phase > 0) is unchanged in both modes.
• parametersVersion() bumped 1 → 2; the three shipping pipelines that use the filter are migrated; SIMPL-converted and older NX pipelines pick up the default via argument backfill.
• Three new inline analytical unit tests (per-voxel multi-feature fixture, two-phase gate fixture, per-grain/per-voxel mode-equivalence invariant) extend the existing Class 1/Class 4 suite — no exemplar archives.
• V&V report reopened (status DRAFT pending re-sign-off): code paths re-enumerated per mode, oracle section extended with the new hand-derived fixtures, and a runtime A/B against DREAM3D 6.5.171 on the default path (identical seeded synthetic input through both runners; max |Δ| = 0.0072°, consistent with the documented EbsdLib 2.4.1 precision deviation). Per-voxel mode is recorded as an NX-only capability deviation.
• Review-driven cleanup of pre-existing code in the algorithm: getDataRefAs for the validated geometry, signed kernel boundary clamps with the neighbor index computed from the clamped indices, LaueOps list built once per run, per-row cancel checks, progress-increment floor, removal of the unreachable numVoxel == 0 fallback, and real Doxygen. All behavior-preserving.

Fixes #1613

Test Plan

• Tests pass on in-core build (filter suite 9/9; OrientationAnalysis 258/258; FilterValidation 5/5)
• Tests pass on out-of-core build (filter suite 9/9)