Expose full tip poses and configuration evolution in the sequential CPU benchmark

[mstoelzle]

Motivation

We want the paper to report more quantitative rollout-accuracy statistics between SoRoSim and SoRoMoX, rather than limiting the comparison to end-effector position error. Full tip poses and configuration trajectories are needed to evaluate orientation and configuration errors throughout each rollout.

Requirements

• Export the full tip pose—including orientation—at every saved timestamp for SoRoMoX, SoRoSim, and PyElastica.
• Export the generalized configuration trajectory (q(t)) for SoRoMoX and SoRoSim. PyElastica’s internal configuration evolution is outside the required scope.
• Document pose conventions, reference frames, units, configuration ordering, and output schemas.
• Update the committed paper data, loaders, and plotting or inspection utilities while retaining the existing position comparison where useful.
• Keep timestamps aligned with the corresponding simulation outputs.

Acceptance criteria

• All four sequential CPU cases provide full tip-pose histories for each backend.
• SoRoMoX and SoRoSim results additionally provide complete configuration histories.
• The exported data supports quantitative comparison of position, orientation, and configuration rollout errors between SoRoSim and SoRoMoX.
• Position trajectories remain recoverable from the new pose data.
• Regenerated canonical artifacts load successfully, and automated tests validate field presence, dimensions, and timestamp consistency.
• The Section IVa README explains how to generate and interpret the additional outputs.