fix: WealthPercentileTransformer silent no-op and dead branch, EncounterTransformer datetime64 rejection

[shivamlalakiya]

Summary

Three real defects found during a pass over the JOSS-prep plan, plus the process fix for how the third one happened.

• WealthPercentileTransformer(wealth_cols=[...]) now raises ValueError instead of silently training without the wealth signal when none of the named columns exist in X. A typo'd column name was indistinguishable from an intentional no-op. Closes WealthPercentileTransformer silently ignores a wealth_cols list that matches no column #156.
• WealthPercentileTransformer.fit's dead hasattr(X, "columns") branch is removed (validate_data already returns a NumPy array and already sets feature_names_in_ for DataFrame input). Two contributors independently fixed this same branch the same evening, fix: remove dead hasattr branch in WealthPercentileTransformer.fit (#168) #175 (@HeaTTap) and fix: remove dead branch in WealthPercentileTransformer.fit (#168) #176 (@Larslllllll); this PR lands that fix and credits both in CONTRIBUTORS.md. Closes cleanup: WealthPercentileTransformer.fit has a branch that can never run #168.
• EncounterTransformer no longer rejects a gift_date column already parsed to datetime64. Any real loader (pd.read_csv(parse_dates=...), a SQL read, this package's own make_donor_panel) hands back exactly that column mix, and validate_data's numpy promotion of datetime64 against float64 raised before the transformer got a chance to read it. Closes EncounterTransformer rejects a parsed datetime64 gift_date and accepts only date strings #163.
• CONTRIBUTING.md gets a "Claiming an issue" rule (comment before you open a PR, wait for assignment) so the fix: remove dead hasattr branch in WealthPercentileTransformer.fit (#168) #175/fix: remove dead branch in WealthPercentileTransformer.fit (#168) #176 duplication doesn't repeat; the issue-draft template states it inline too.

Also filed 5 good-first-issue drafts from scripts/issue-drafts/ as real GitHub issues (#181-#185), each re-verified against current line numbers/greps first. Not part of this diff since they're process, not code.

Test plan

• New regression tests for all three fixes: test_unmatched_wealth_cols_raises, test_feature_names_match_for_frame_and_array_input, test_accepts_parsed_datetime64_gift_date
• make ci green
• make riskcov green (97% branch over the risk-tier subtree, 93% floor)
• Full suite green

[shivamlalakiya]

Closing this in favor of already-open external contributor PRs for the same issues: #180 (@be-student, issue #163) and #179 (@be-student, issue #156, which also removes the #168 dead branch). Those landed first, are already green, and external contribution signal is a tracked JOSS blocker, so they should merge over a duplicate from me.

The parts of this PR not covered by #179/#180 (the CONTRIBUTING.md "Claiming an issue" rule and the issue-draft template line) are being handled separately to avoid a third duplicate.