feat: scaffold Week 12 assignment (Airflow orchestrated pipeline)

[lassebenni]

Summary

Scaffolds the Week 12 Airflow orchestrated-pipeline assignment (curriculum Ch11) as an Astro project. Students implement dags/taxi_pipeline.py: a scheduled DAG with catchup=False + a tag, three sequential tasks (ingest_taxi_month → dbt_run → dbt_test) where dbt runs via uvx --python 3.11 (plain dbt crashes on the image's Python 3.14), logical-date partitioning, and retries. Provided: the Ch6 tests/test_dag_integrity.py, a Dockerfile (Astro Runtime 3.3), requirements.txt (providers, no dbt), and RUNBOOK / ASSIGNMENT_REPORT / AI_ASSIST templates.

⚠️ Supersedes PR #1 for this repo. PR #1 scaffolds a Metabase + Streamlit dashboarding assignment, which does not match the current Week 12 (Airflow) curriculum. That dashboarding scaffold belongs to a different week — please redirect or close #1.

Autograder verified end-to-end

.hyf/test.sh is a static 100-pt ladder (pass = 60), reusing grader_lib.sh. No Airflow install needed — pure bash/sed/grep, so it runs in the shared HYF auto-grade CI.

Score	Gate
20	Required files exist
+15	DAG implemented (no NotImplementedError; @dag + tasks)
+20	Three tasks ingest → dbt_run → dbt_test, chained with >>
+20	dbt via uvx; retries configured
+15	Logical-date partition ({{ ds }}/logical_date); catchup=False
+10	RUNBOOK.md + AI_ASSIST.md filled (no leftover TODO)

Test plan

• Bare scaffold: 25/100, pass=false (only files-exist + DAG structure)
• Full solution (snapshot DAG + filled docs): 100/100, pass=true
• Stub + solution DAGs py_compile clean; test.sh bash -n clean
• Guidance kept out of grep-target literals (docstrings/comments don't pre-satisfy checks)

🤖 Generated with Claude Code