Replace Jinja2 prompt templating with stdlib prompt composition

[Cataldir]

Context

Jinja2 is currently imported at service startup, which is causing local failures. It is considered overkill for current prompt needs.

Scope

• Essays, questions, and upskilling prompt composers and prompt files.
• Remove scoped Jinja2 dependencies and mentions in documentation.

Acceptance Criteria

• Use stdlib or simple slots for templating.
• Dynamic sections should be built in Python.
• Include tests for prompt rendering.
• No changes to the tutor_lib.agents boundary.
• Documentation must be updated.

Risks

• Prompt drift.
• Whitespace issues.
• Missing placeholders.

Architecture Review

Remove Jinja2; use lightweight Builder-style prompt block functions, not Template Method inheritance.

[Cataldir]

Implementation completed in workspace.

Summary:

• Replaced Jinja2 prompt composition in essays, questions, and upskilling with stdlib string.Template and Python Markdown block builders.
• Converted scoped prompt artifacts from .jinja to .md.
• Removed direct Jinja2 dependencies from the scoped service pyproject.toml files.
• Updated architecture docs from Jinja2-specific wording to lightweight stdlib prompt composition.
• Added focused prompt-composition tests.

Validation:

• Targeted prompt tests: 40 passed.
• Full test suite: 148 passed, 23 warnings.
• Active Python still has no jinja2 installed, and apps/**/*.jinja count is 0.
• VS Code launch.json validates.

Known residual note:

• VS Code diagnostics still show pypdf.errors unresolved in the editor environment for essays.py, but pytest is green and pypdf is already declared by the essays service.