Issue #20: bimodal winner distributions, 3-2-1, partisan primaries, Score/STAR ranges

[endolith]

This PR implements the simulations and mechanisms discussed in issue #20.

Library additions

• bimodal_electorate (elsim.elections): two voter clusters and matching candidate pools at ±separation on the first axis (documented indexing convention for left/right voters and candidates).
• three_two_one (elsim.methods): 3-2-1 voting with ballots encoded as 0 = Bad, 1 = OK, 2 = Good (party-specific semifinalist rules omitted as optional for simulations).
• honest_321_ratings (elsim.strategies): per-voter tertiles → Good/OK/Bad.
• honest_normed_scores(utilities, max_score=5, *, min_score=0): keyword-only min_score; keeps positional max_score for backward compatibility; rows with no utility spread map to min_score.
• Partisan helpers (elsim.methods): restricted plurality nominees, pairwise majority between finalists, open vs closed partisan flows, top-two runoff with a subset electorate in the pairwise round.

Examples (same style as distributions_by_method.py)

• examples/bimodal_winner_distributions.py: 5k+5k voters, two clusters of candidates, dispersion 0.5, reduced-turnout primaries (~3.8k each side) and runoff (~8.5k voters). Methods include FPTP, contingent runoff, IRV, Score/STAR (0–5), Black, open/closed partisan primaries, 3-2-1, approval (mean threshold and vote-for-half), reference histograms.
• examples/score_star_range_comparison.py: same unimodal spatial setup as existing 1D plots; compares Score vs STAR for ranges 0–5, 0–10, 1–10, five-grade (0–4), 13- and 15-point scales.
• examples/bimodal_approval_strategy_comparison.py: optimal approval vs vote-for-half vs vote-for-one on the bimodal electorate.

Tests

• bimodal_electorate shape/means; three_two_one edge cases; extended invalid tiebreaker coverage.

All tests pass (pytest tests/).

  

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.23%. Comparing base (3249c6c) to head (8af7090).

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #50      +/-   ##
==========================================
+ Coverage   96.37%   97.23%   +0.86%     
==========================================
  Files          17       19       +2     
  Lines         496      652     +156     
==========================================
+ Hits          478      634     +156     
  Misses         18       18              

Flag	Coverage Δ
no-numba	96.77% <100.00%> (+1.01%)	⬆️
numba	90.95% <100.00%> (+2.84%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.