Make above-the-line deduction sums deterministic

[MaxGhenis]

Why

adjusted_gross_income_person and student_loan_interest_ald_magi build their list of tax-unit deductions as list(set(all_alds) - set(PERSON_ALDS)). Set iteration order over strings follows the process hash seed, so the float32 accumulation order, and therefore the last digits of AGI, could differ between processes. The canonical SPM acceptance matrix's seed-invariance (byte-identical vectors under PYTHONHASHSEED 0 and 42) rests on the accepted candidate wheel having this fix; the fix lived on the determinism branch and never reached the merged #9428, so the published 2.0.0 sums in hash-seed order.

What

• sorted(...) in both formulas (two lines).
• policyengine_us/tests/core/test_ald_determinism.py: both formulas, with and without deductions, in both list orders, plus a subprocess check under two hash seeds.
• Changelog fragment (fixed).

The candidate's exact policyengine-core==3.32.5 pin is deliberately not ported: the wrapper bundle pins core exactly; the country keeps >=3.30.1.

Verification

uv run python -m pytest policyengine_us/tests/core/test_ald_determinism.py: see the run recorded below the fold in the rollout runbook; ruff format check clean.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 62.29%. Comparing base (f9b5549) to head (064c3c9).
⚠️ Report is 69 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff              @@
##              main    #9446       +/-   ##
============================================
- Coverage   100.00%   62.29%   -37.71%     
============================================
  Files            2        2               
  Lines           20       61       +41     
  Branches         0        3        +3     
============================================
+ Hits            20       38       +18     
- Misses           0       22       +22     
- Partials         0        1        +1     

Flag	Coverage Δ
unittests	62.29% <100.00%> (-37.71%)	⬇️

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.