Fix BUG-004/005: mixed-materials load accumulation and plane-stress postprocessing

[knutambot]

Summary

• BUG-004: fext accumulated across time steps in the mixed-materials tutorial (never cleared before apply!(fext, lh, t)), so the solver applied the sum of all previous loads instead of the current one. Fixed with fill!(fext, 0) before each step's apply!.
• BUG-005: the tutorial's calculate_stress postprocessing zero-padded the out-of-plane strain and applied the full 3d elastic stiffness directly, which is only valid for plane strain, not plane stress (whose eliminated out-of-plane strain must satisfy σ33 = 0). Fixed by delegating to MaterialModelsBase.material_response on the actual (stress_state, material) pair for the elastic response, and by using the already-converged 3d plastic strain (reduced to in-plane) to form the elastic strain for the plastic branch — without re-invoking Plastic's own material_response a second time, which would otherwise double-advance the already-converged state during postprocessing.
• Added hidden (#src) regression checks: per-timestep fext reconstruction check, an elastic plane-stress analytical reference (E/(1-ν²) form) plus an out-of-plane-stress check, and a pinned end-to-end regression value for the full solve's postprocessed stresses.

Verified the corrected formula numerically against MaterialModelsBase.material_response directly: for E=210e3, ν=0.3, ϵ11=0.01, it gives (σ11, σ22) = (2307.69, 692.31), matching the documented correct reference (vs. the buggy (2826.92, 1211.54)).

Test plan

• Pkg.test() — full suite passes (3236 assertions, no failures)
• All tutorials (heat_equation, iga, incompressible_elasticity, mixed_materials, viscoelasticity) run successfully, each in an isolated module
• All how-tos (automatic_differentiation, local_constraints, robin_bc, surface_integral, threaded_assembly, volume_integral) run successfully
• Full docs/make.jl build succeeds (only the pre-existing benign viscoelasticity image-size-threshold warning)
• Independent Codex review of the plan and of the final diff (same review thread)

Remaining risk

Codex's review flagged (medium, both passes) that docs/src/literate_tutorials/mixed_materials.png is a downloaded external asset (via FerriteAssembly.asset_url, not tracked in this repo / not regenerated by the docs build), so this PR does not update the published tutorial image — it will keep showing results from the old, incorrect loading/postprocessing until that asset is separately regenerated and republished upstream.

🤖 Generated with Claude Code

https://claude.ai/code/session_01KySgpVgJ5fWR9AvPK1JQoU

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.17%. Comparing base (d30c676) to head (0111539).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #94   +/-   ##
=======================================
  Coverage   97.17%   97.17%           
=======================================
  Files          32       32           
  Lines        1347     1347           
=======================================
  Hits         1309     1309           
  Misses         38       38           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.