[evandhoffman]

Closes #38.

What changed

• enrich repeated-run model sections with wall-time variance, p90, distinct app-hash counts, distinct solution-shape counts, and median timing breakdowns
• expand the repeated-run summary table so variance and stability signals stay visible at a glance
• add an order-effects table that summarizes pass rate and median wall time by run position across rounds
• keep single-run report output unchanged

Verification

• uv run python -m unittest tests.test_report
• uv run python -m unittest discover tests