test(models): add smoke test that every DH/URDF/ETS model constructs

[petercorke]

Summary

• tests/test_models.py is entirely hand-written, one method per model — so a model exported in __all__ can go completely untested. Valkyrie, Fetch, KinovaGen3, FetchCamera, and LBR all had zero coverage this way; three of them were silently broken with nothing to catch it (see tech-debt.md, and fix(models): patch broken upstream RD files for Valkyrie and Fetch #543/fix(models): remove FetchCamera, unshippable and unused in RVC3 #545/fix(models): make KinovaGen3 load via robot_descriptions, add XACRO_ARGS support #546).
• Adds a generic TestModelSmoke.test_all_models_construct that iterates each of DH/URDF/ETS's own __all__ (not inspect.getmembers/module __dict__ scanning, which would also pick up re-exported base classes like Robot/URDFRobot) and asserts every listed class constructs with no arguments.
• Known-currently-broken models are tracked in an explicit EXPECTED_FAILURES set. This fails loudly in both directions: a newly-broken model fails the test, and a since-fixed "expected failure" that starts passing also fails the test — forcing whoever fixes it to clean up the entry rather than letting it go stale.
• EXPECTED_FAILURES currently lists 5 entries: LBR (needs an rtb-data release, tracked in tech-debt.md) plus Valkyrie/Fetch/KinovaGen3/FetchCamera, all already fixed on unmerged branches (fix(models): patch broken upstream RD files for Valkyrie and Fetch #543, fix(models): remove FetchCamera, unshippable and unused in RVC3 #545, fix(models): make KinovaGen3 load via robot_descriptions, add XACRO_ARGS support #546) — each entry documents which PR resolves it, for easy cleanup once merged.

Test plan

• Verified on clean main: 29/29 pass in test_models.py, only the 5 documented models hit EXPECTED_FAILURES
• Sanity-checked the mechanism actually works: temporarily cleared EXPECTED_FAILURES and confirmed the test fails, listing exactly the 5 known-broken models with their real exceptions — nothing else
• Full suite: 653 passed, 13 skipped, no regressions

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 0.00%. Comparing base (17c67d2) to head (553c93f).

Additional details and impacted files

@@          Coverage Diff          @@
##            main    #547   +/-   ##
=====================================
  Coverage   0.00%   0.00%           
=====================================
  Files        136     136           
  Lines      13438   13438           
=====================================
  Misses     13438   13438           

☔ 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.