Align local grading and assertion descriptions with backend behavior

[skovranek]

• Align bootdev local-test with backend grading so local results reflect submission behavior.
• Evaluate every assertion in each test, require one result per step, and align command failures, JSON integer comparisons, response captures, and header/trailer comparisons.
• Header names remain case-insensitive; values are case-sensitive.
• Match backend interpolation rules and update assertion descriptions and jq previews accordingly. Validate malformed assertions and ambiguous capture definitions before running steps.