[QTI] Dual-publish Perseus zips for choice/text-entry (math input only) QTI items

[rtibbles]

❌ This issue is not open for contribution. Visit Contributing guidelines to learn about the contributing process and how to find suitable issues.

Overview

Derive a Perseus archive from QTI items whose interaction type Perseus can express (single/multiple choice, text/math input), published alongside the QTI package so older Kolibri still renders the content.

Complexity: High
Target branch: unstable

Context

• Older Kolibri renders Perseus; newer renders QTI. The authored source of truth is QTI XML in raw_data.
• PerseusExerciseGenerator builds Perseus from structured question/answers, which native QTI items no longer carry.
• The pydantic models can read the choice/text-entry subset. Advanced interaction types have no Perseus equivalent.

The Change

• For QTI items mapping to a Perseus question type, derive the structured form from the QTI XML and produce a Perseus archive in addition to the QTI package.
• Publish QTI only for nodes containing any interaction with no Perseus equivalent — never a partial or invalid Perseus.

Acceptance Criteria

• A node of choice/text-input QTI items publishes both a QTI package and a Perseus archive
• The derived Perseus matches the QTI item: prompt, choices, correct answers
• Nodes containing any non-Perseus-expressible interaction publish QTI only
• Older Kolibri (Perseus) and newer Kolibri (QTI) each render the node
• Tests cover derivation for single choice, multiple choice, and text/math input, plus the QTI-only fallback

References

• contentcuration/contentcuration/utils/assessment/perseus.py

AI usage

Architecture decided with the maintainer across an iterative session: blanket QTI type with the item XML in raw_data; XSD-authoritative validation across all sources; a legacy→QTI global migration with an API-layer dual-read; ricecooker upload delegating to the AssessmentItem serializer; and a Perseus custom-interaction contract confirmed against the QTI 3.0 specification. Claude mapped the existing publish/validation/ricecooker code, proposed the breakdown, and drafted each issue; the maintainer steered every decision and reviewed throughout.