Move the two-child limit child-poverty comparison from the validation page to Scorecard #2704

docs/book/validation/child-poverty-tcl.md hard-codes PolicyEngine UK's estimates of removing the Universal Credit two-child limit beside the published DWP, HM Treasury and Resolution Foundation figures. #1724 corrected those numbers by hand. They go stale with every model or data release, and the page cannot say for each row whether PolicyEngine's concept matches the publisher's.

PolicyEngine Scorecard now holds the comparison: PolicyEngine/policyengine-scorecard#225 recomputes it on a certified policyengine.py release and annotates each row's concepts. These include:

• the relative-line convention;
• the absolute line's reference year;
• AHC or BHC;
• the child count;
• the financial-year mapping;
• whether the cost is static or post-behavioural.

Each row also keeps its earlier runs in its history.

What done looks like

• The page keeps the method, meaning the reform, the variables and a policyengine.py snippet. It links each benchmark to its Scorecard row and drops the hard-coded results.
• Pages that describe it say where the comparison now lives: reform-impacts-dashboard-plan.md and two-child-limit-full-rollout.md.

This is separate from #1398, which tracks baseline child-poverty benchmarks and stays open.