Add higher-order CommonPlane quadrature rows

[ebchin]

Context

This PR builds on #248. The first PR preserves each redecomposed face's map to its native parent face; this PR uses that map to evaluate higher-order geometry and fields at CommonPlane integration points without treating the LOR finite-element space as the source of the physical solution.

Summary

• Add configurable single-point, multipoint, and order-aware automatic integration for CommonPlane overlaps.
• Build reusable device-resident CommonPlane quadrature rows from accepted LOR overlap geometry.
• Map each row to both native parent faces for position, velocity, and basis evaluation, then scatter explicit penalty forces directly to native parent degrees of freedom without LOR field transfer.
• Refresh native parent coordinate, velocity, inverse-mass, and response data only for CommonPlane schemes.
• Keep connectivity in the requested execution memory space while using explicit host mirrors of element maps for MFEM's host-only record and response transfers.
• Use the same row data for explicit force evaluation and preserve existing low-order CommonPlane behavior where its original integration is exact.

Validation

• Built all affected targets in the release configuration with warnings treated as errors.
• Passed tribol_common_plane_penalty, tribol_enforcement_options, and tribol_iso_integ.
• Passed the full two-rank tribol_mfem_common_plane_2rank and tribol_mfem_mortar_lm_2rank regressions.
• Coverage includes polynomial exactness, 2D and 3D multipoint integration, localized warped-face contact, native parent-field evaluation, direct parent response scatter, all supported face mapping types, no-LOR identity mapping, and LOR factors two and three.
• Passed clang-format-19 --dry-run --Werror and git diff --check.

Stack

• Depends on Store native parent-face mappings for LOR CommonPlane contact #248.
• Base branch: feature/ebchin/common-plane-lor-geometry.