runoff: reuse IRV Numba tally helpers for head-to-head round

[endolith]

Closes #3.

Summary

The contingent-vote / top-two runoff second round used a Python loop with list.index per ballot. That round is equivalent to instant runoff among only the two finalists: skip every other candidate on each ballot, then count first preference among {finalist_0, finalist_1}. That is exactly what _inc_rank_idx + _tally_at_rank_idx do—the same Numba path IRV and Coombs use.

Changes

elsim/methods/runoff.py

• Replace the per-ballot list.index head-to-head tally with _inc_rank_idx / _tally_at_rank_idx.
• Build an eliminated_mask (all candidates except the two finalists) and advance each voter's rank cursor past eliminated candidates before tallying.
• Uses the same boolean-mask API as IRV/Coombs after master dropped reflected Python set in Numba helpers.

tests/test_runoff.py

• test_docstring_example: exercises the contingent-vote example from the runoff() docstring.
• test_head_to_head_tally_via_irv_helpers: asserts the shared helpers produce the expected 3–2 split when only finalists A and B remain.

Rebase notes

Rebased onto current master. The original PR passed a Python set into _inc_rank_idx; master now requires a boolean eliminated_mask ndarray, so the second commit updates runoff accordingly.

Testing

• pytest: 217 passed locally (with [fast] / Numba).
• Existing runoff property tests (Hypothesis) unchanged and still pass.

  

[what-the-diff]

PR Summary

• New Import Functions Introduced: The code now includes two new import functions called _inc_rank_idx and _tally_at_rank_idx from the _common module. These tools help streamline the process of updating ranks and tallying votes.

• Improved Vote Counting Process: The logic behind counting votes for two options, finalist_0 and finalist_1, has been refined. It now uses a method from the library NumPy, which simplifies calculations and makes the process more efficient than the former looping method.

• Introduction of eliminated_cands Set: A new set, eliminated_cands, was added. This serves as a check list to keep track of candidates who are no longer under consideration.

• Maximized use of New Methods: The newly added functions _inc_rank_idx and _tally_at_rank_idx are now being utilized for updating ranking index and tallying votes respectively. This change was made to make the overall workflow more efficient and reliable.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.39%. Comparing base (3249c6c) to head (4b7478a).

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #46      +/-   ##
==========================================
+ Coverage   96.37%   96.39%   +0.02%     
==========================================
  Files          17       17              
  Lines         496      499       +3     
==========================================
+ Hits          478      481       +3     
  Misses         18       18              

Flag	Coverage Δ
no-numba	95.79% <100.00%> (+0.02%)	⬆️
numba	88.17% <100.00%> (+0.07%)	⬆️

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.