fix(rne): propagate backward-pass force through the correct transform direction

[petercorke]

Problem

Robot.rne()'s backward recursion (Robot.py) propagated force through
Xup[j], which is the child<-parent motion transform (used in the forward
pass: v_child = Xup[j] * v_parent). But force needs to propagate the
other direction, child->parent, and needs the transform in that direction
instead.

This was masked by a bug in spatialmath-python's SE3 * SpatialForce,
which used the adjoint transpose instead of the coadjoint (fixed upstream in
spatialmath-python 1.1.18) —
the old, incorrect formula happened to cancel out RTB's own direction
mismatch for this specific operation. With the upstream fix, that masking
went away: Robot.rne() (used by all URDF/ETS-based robots) started
returning wrong torques for any joint with a translational offset from its
parent — i.e. virtually every real robot. DHRobot.rne()'s compiled C++
path never touches spatialmath's SE3/SpatialForce classes, so it was
unaffected throughout — which is how this got isolated.

Fix

Propagate force through the un-inverted transform, Xup[j].inv() (i.e. the
parent<-child direction), matching the direction spatialmath's corrected
SE3 * SpatialForce now expects.

Verification

• DHRobot / Puma560 reference torques (compiled C++, independent of
  spatialmath's Python classes) — unaffected throughout, used as ground
  truth
• test_invdyn — Spong textbook 2-link closed-form — now passes under
  spatialmath-python 1.1.18 (previously failed)
• test_symdyn — symbolic RNE — expected string regenerated; Q[1] is
  byte-identical to before, Q[0] differs only by an unsimplified
  sin(q1)**2 + cos(q1)**2 term introduced by the added .inv()
  (confirmed algebraically/numerically equivalent to the old expression)
• 6x test_fknm_fallback.py — now pass
• Full local suite: 735 passed, 82 skipped, 0 failed
  (pytest tests/ --ignore=tests/test_blocks.py, spatialmath-python 1.1.18)

Also bumps pyproject.toml's spatialmath-python floor to >=1.1.18,
since that's the first release where SE3 * SpatialForce is correct.

Test plan

• Full test suite passes locally against spatialmath-python 1.1.18
• CI green

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 0% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 0.00%. Comparing base (c040522) to head (b8ee0d6).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/roboticstoolbox/robot/Robot.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@          Coverage Diff          @@
##            main    #691   +/-   ##
=====================================
  Coverage   0.00%   0.00%           
=====================================
  Files        143     143           
  Lines      14082   14082           
=====================================
  Misses     14082   14082           

☔ 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.