feat: implement using multiple stimuli versions

[cbueth]

References

• PR feat: Implement using multiple stimuli versions #63
• Branch: based on the existing PR, but needs to be redone on current main

Goal

The MultiplEYE experiment uses multiple versions of stimuli (different orderings, different answer option shuffling). The preprocessing pipeline needs to handle these versions correctly so that reading measures and answer keys are matched to the right stimulus version per participant.

Status quo

PR #63 exists as a first attempt but it is based on an older codebase state. The current pipeline does not fully account for multiple stimulus versions when computing reading measures or mapping answer keys.

What needs to change

• Rework the stimulus version handling from scratch, using PR feat: Implement using multiple stimuli versions #63 as reference but rebased on current main after the output restructuring refactor!: restructure output folders #98.
• The stimulus version determines: ordering of trial presentation, which answer options map to which directional keys, which text/stimulus appears in each trial slot.
• Version information is stored in files under config/ (e.g. question_order_versions.csv, question_answer_option_shuffling_*).
• The pipeline must read the version for each session and pass it through to trial-level processing: reading measures computation, answer parsing, and any per-trial metadata.

Grains of salt

• This depends on refactor!: restructure output folders #98 and issue feat: parse and store comprehension question answers #100 being done first, since version info affects both reading measures and answer mapping.
• The versioning logic is spread across config files and experiment data. Make sure to understand how versions are assigned per session before designing the solution.

Test strategy

Run pipeline on two sessions with different stimulus versions. Verify trial order, stimulus mapping, and answer key mapping differ correctly.