fix(ik): _random_q() silently produces garbage for non-finite joint limits

[petercorke]

Summary

RTB has two separate IK implementations reachable from different public method names:

• ikine_LM/ikine_NR/ikine_GN/ikine_QP -- the pure-Python IK_LM/IK_NR/IK_GN/IK_QP classes in IK.py.
• ik_LM/ik_NR/ik_GN -- documented as "a fast solver implemented in C++", backed by ik.cpp via nanobind.

Both had the identical gap: their random-restart sampling (IK.py's _random_q() and ik.cpp's own _rand_q()) sampled directly from a joint's qlim with no check that the limits were actually finite. A joint with a bad (non-finite) limit baked into its own model data -- this was the real root cause behind #485's KinovaGen3 report, whose own qlim was fixed separately without ever patching either solver -- caused:

• Python path: a silent NaN joint value, or an opaque internal numpy error (OverflowError: high - low range exceeds valid bounds), depending on which RNG code path is hit.
• C++ path: silently returned a "solution" with q=[nan], success=0, after burning through every one of the 100 random restarts -- no exception, no diagnostic at all.

Since RTB's public API is IK regardless of which implementation backs a given method name, both are fixed here, in lockstep.

Fix

• IK.py's _random_q(): checks np.isfinite() on the joint limits before sampling, raises ValueError naming the offending joint index(es) and their bad qlim.
• ik.cpp's _rand_q(): same check via std::isfinite, throws nb::value_error (maps to a Python ValueError) before sampling.

Both match the existing convention elsewhere in this code (e.g. ETS.qlim already raises for an unset prismatic limit) of failing loudly rather than propagating a silently-bad value.

Test plan

• New regression test test_random_q_rejects_non_finite_qlim (Python path): a joint with qlim=[-inf, inf] raises ValueError; a normal finite-limit joint is unaffected. Confirmed fails (opaque OverflowError) against pre-fix code, passes with the fix.
• New regression test test_ik_lm_c_rejects_non_finite_qlim (C++ path): ets.ik_LM() with the same bad joint raises ValueError. Verified by rebuilding the compiled extension locally and confirming fail-then-pass directly -- pre-fix, ets.ik_LM() silently returned (q=[nan], success=0) after 101 wasted searches; post-fix, raises immediately.
• Full test suite green: 640 passed, 75 skipped (unrelated optional-dep skips), no regressions -- run against the freshly rebuilt extension.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 0% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 0.00%. Comparing base (5b5a0fe) to head (562566f).

Files with missing lines	Patch %	Lines
src/roboticstoolbox/robot/IK.py	0.00%	6 Missing ⚠️

Additional details and impacted files

@@          Coverage Diff          @@
##            main    #648   +/-   ##
=====================================
  Coverage   0.00%   0.00%           
=====================================
  Files        143     143           
  Lines      14035   14039    +4     
=====================================
- Misses     14035   14039    +4     

☔ 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.