Qti exercise export

[rtibbles]

Summary

• Adds Pydantic models to allow validated construction of a subset of the QTI XML schema
• Adds Pydantic models to allow validated construction of HTML
• Adds Pydantic models to allow validated construction of MathML
• Adds utilities for parsing assessment item markdown to HTML and MathML for spec compliant bundling in QTI
• Refactors Perseus exercise archive generation into an abstract base class and a specific perseus class
• Creates a correlated QTI archive generation using the same base class
• Makes creation of Perseus or QTI conditional on the existence of Free Response items in the AssessmentItems
• Adds copious tests for most of the above.

References

Fixes #4878

Reviewer guidance

Creating a "Survey" exercise with a free response question and then checking you can successfully publish it is the main manual check.

The generated QTI item XML can be cross checked against https://qti.amp-up.io/testrunner/sandbox/

The IMSManifest format should be checked against the examples here: https://www.imsglobal.org/spec/qti/v3p0/guide#h.r1pkoq891wk7 and here: https://www.imsglobal.org/spec/qti/v3p0/impl#h.5esb962k7g83

The QTI schema is incredibly verbose. Where I can, I have attempted to slightly reduce the verbosity if possible, but mostly the verbosity of the schema has resulted in a lot of structural code.