fix(control): clear completed JTT state and remove idle holding

[TomCC7]

Contribution path

Fixes a regression introduced in #3610.

Problem

• JTT reused stale commanded positions after another task moved the joints, causing dangerous jumps on the next execution.
• Idle holding duplicated the shared hardware interface's handling of retained joint targets.

Solution

• Clear each joint's commanded state after its final output; initialize new executions from measured positions.
• Keep commanded-state continuity during active execution and trajectory replacement.
• Remove hold_position_when_idle and its G1 configuration.
• Add regression tests verifying JTT does not pull joints back after another task moves them.

How to Test

• Manually verified on hardware: teleop followed by trajectory execution no longer produces the dangerous motion.
• Verified trajectory completion and teleop handoff in G1 MuJoCo simulation.
• 128 unit tests passed:

uv run pytest dimos/control/tasks/trajectory_task/test_trajectory_task.py dimos/control/test_control.py dimos/control/test_coordinator_routing.py -q

AI assistance

Codex (GPT-6): diagnosis, implementation, tests, and PR description.

Checklist

• I have read and approved the CLA.

[codecov]

Codecov Report

❌ Patch coverage is 98.48485% with 1 line in your changes missing coverage. Please review.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
...trol/tasks/trajectory_task/test_trajectory_task.py	98.30%	0 Missing and 1 partial ⚠️

@@            Coverage Diff             @@
##             main    #3969      +/-   ##
==========================================
- Coverage   77.91%   77.91%   -0.01%     
==========================================
  Files        1398     1399       +1     
  Lines      132268   132293      +25     
  Branches    11487    11484       -3     
==========================================
+ Hits       103061   103079      +18     
- Misses      25943    25949       +6     
- Partials     3264     3265       +1     

Components	Coverage Δ
Tests	94.72% <98.30%> (-0.01%)	⬇️

Flag	Coverage Δ
OS-ubuntu-24.04-arm	74.26% <98.48%> (+<0.01%)	⬆️
OS-ubuntu-latest	74.76% <98.48%> (+<0.01%)	⬆️
Py-3.10	74.76% <98.48%> (+<0.01%)	⬆️
Py-3.11	74.75% <98.48%> (+<0.01%)	⬆️
Py-3.12	74.75% <98.48%> (+<0.01%)	⬆️
Py-3.13	74.76% <98.48%> (+<0.01%)	⬆️
Py-3.14	74.76% <98.48%> (+<0.01%)	⬆️
Py-3.14t	74.76% <98.48%> (+<0.01%)	⬆️
SelfHosted-Large	30.71% <18.18%> (-0.01%)	⬇️
SelfHosted-Linux	35.76% <19.69%> (-0.02%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...s/control/tasks/trajectory_task/trajectory_task.py	89.20% <100.00%> (+0.51%)	⬆️
dimos/control/test_control.py	96.74% <ø> (-0.10%)	⬇️
...nitree/g1/blueprints/basic/unitree_g1_groot_wbc.py	58.33% <100.00%> (+0.34%)	⬆️
...trol/tasks/trajectory_task/test_trajectory_task.py	98.30% <98.30%> (ø)

... and 6 files with indirect coverage changes

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[greptile-apps]

Greptile Summary

No merge-blocking issue was established in the changed trajectory control logic.

T-Rex validation blocked

The focused trajectory lifecycle check and narrow control suite could not reach the changed code because the temporary environment was missing dimos_lcm and pydantic_settings.

Confidence Score: 5/5

No established issue blocks merging.

No severity-bearing findings were established.

Files Needing Attention: No specific file requires follow-up; the focused control suite should be rerun in an environment with the required Python dependencies.

T-Rex Logs

What T-Rex did

• A focused trajectory lifecycle harness, including a trajectory factory, completion, replacement, and preemption, was prepared for before-and-after execution.
• The base harness import attempt halted during imports because dimos_lcm was unavailable, preventing any control behavior from running.
• The updated harness import attempt halted during imports for the same dimos_lcm blocker, before pre-execution work occurred.
• The narrow control suite loading stopped while loading test configuration because pydantic_settings was unavailable, preventing test collection.
• The harness and validation runner artifacts were uploaded for review, including the harness source and the run-validation script.

_{Ran code and verified through T-Rex}

_{Reviews (1): Last reviewed commit: "test(control): assert JTT continuity aft..." | Re-trigger Greptile}