Refactor Akinci surface tension foundations

[svchb]

Summary

• add dispatch-based surface-normal requirements so CohesionForceAkinci no longer allocates or computes unused normals
• validate all surface-tension coefficients and stabilize Akinci wall adhesion near the support boundary
• promote mixed ColorfieldSurfaceNormal threshold types and skip the Morris capillary restriction for a zero coefficient
• document the dimensional limitations of Akinci kernels in 2D
• add cohesion-only and wetting examples, and make maintained surface-tension examples use valid configurations

Stack

This is PR 1 of the surface-tension refactoring stack. It targets bug_fixed. Follow-up PRs will address Morris geometry/CSF, CSS interaction conservation and diagnostics, then model comparison and validation cases.

Findings addressed

• C08/C32: remove unnecessary normal setup for cohesion-only Akinci
• C09: avoid the near-support adhesion DomainError
• C17: support mixed numeric normal thresholds
• C19/C31: reject invalid surface-tension coefficients
• C27: define and document the current 2D Akinci dimensional policy
• C07/C28: keep maintained examples constructible while the final interaction matrix is developed

Verification

• surface-tension tests: 82/82
• surface-normal tests: 375/375
• semidiscretization tests: 17/17
• WCSPH tests: 224/224
• Aqua passed
• documentation build and doctests passed
• JuliaFormatter and git diff --check passed
• new examples, wetting presets, falling-spheres model matrices, oil-film, wall, and dam-break overrides passed
• focused examples retained zero RHS allocations; cohesion also passed with four Julia threads

Notes

The oil-film example keeps surface tension disabled until the explicit two-phase interaction policy is implemented in PR 3. The published Akinci normalization is retained for compatibility and documented as empirical in 2D.

[svchb]

Superseded by trixi-framework#1253, which targets upstream main.