Add a leadership annual-giving upgrade notebook

[shivamlalakiya]

Summary

• Adds examples/notebooks/05_leadership_upgrade.ipynb: builds an upgrade training panel with build_upgrade_snapshots on top of make_donor_panel, plus a small synthetic activity log, fits MajorGiftClassifier, and validates it with FiscalYearGroupedSplitter's walk-forward split so no future fiscal year leaks into training.
• Compares the model's top 20% by affinity score against a naive gave-500-plus-last-FY rule on the same out-of-fold predictions: the model's top 20% upgrades at roughly 21% versus about 15-16% for either version of the rule and 12% overall. The notebook states plainly, in its own markdown cell, that this is a smoke test on synthetic data and not a benchmark claim.
• Lists the notebook in examples/README.md's notebook table and adds a CHANGELOG.md entry under Unreleased.

Test plan

• python -m pytest --nbmake examples/notebooks -q --no-cov (all five notebooks, 01-05, re-executed): 5 passed
• make ci: 2177 passed, 8 skipped, coverage 98.62% percent (floor 92 percent)
• make riskcov: risk-tier coverage 98 percent (floor 93 percent)

Note: this branch was originally meant to stack on e4-upgrade-snapshots (#242), but that PR merged into main while this one was in progress, so it is rebased onto current main and opened against main directly instead.