feat: add council-tax single-person discount

[vahid-ahmadi]

Summary

Closes part of #42. Households with exactly one adult (18+) now receive a 25% single-person discount on the calculated council tax — Local Government Finance Act 1992 s.11(1)(a).

sim_couple = Simulation.from_situation({
    "people":     {"p1": {"age": 35}, "p2": {"age": 33}},
    "benunits":   {"b": {"members": ["p1", "p2"]}},
    "households": {"h": {"members": ["p1", "p2"], "main_residence_value": 80_000}},
}, year=2025)
# baseline_council_tax_calculated → £2,280 (Band D, no discount)

sim_single = Simulation.from_situation({
    "people":     {"p": {"age": 35}},
    "benunits":   {"b": {"members": ["p"]}},
    "households": {"h": {"members": ["p"], "main_residence_value": 80_000}},
}, year=2025)
# baseline_council_tax_calculated → £1,710 (= £2,280 × 0.75)

What's included

Engine (src/parameters/mod.rs, src/variables/wealth_taxes.rs, src/engine/simulation.rs):

• New single_person_discount_rate field on CouncilTaxParams (default 0.25)
• calculate_council_tax(hh, params, is_single_adult) applies the discount when the flag is true
• simulation.rs counts adults via Person::is_adult() (age >= 18) and passes adult_count == 1

New microdata columns: baseline_council_tax_calculated and reform_council_tax_calculated on each household row, so reform analyses can isolate the change without decomposing total_tax

Python wrapper: first-time exposure of CouncilTaxParams (the params struct existed in Rust but was never reformable from Python). Reform authors can now do:

from policyengine_uk_compiled import Parameters, CouncilTaxParams
reform = Parameters(council_tax=CouncilTaxParams(single_person_discount_rate=0.50))

Rust unit tests (3): Band D discount, Band A discount, zero-discount-rate edge

YAML policy-test cases (4) in tests/policy/council_tax.yaml: Band D two adults, Band D single adult, Band A single adult, Band D one adult + two children (children don't count for the discount)

Changelog fragment under changelog.d/added/

Caveat — baseline runs are unchanged

Net income still uses the FRS-recorded hh.council_tax amount, which already reflects whatever discount the household actually receives. The calculated value is for reform modelling. Wiring the calculated value into actual net-income computation is a follow-up — it'd require correctly handling Council Tax Reduction (CTR), exemptions, and devolved CTR variants (issues remaining under #42).

Verified locally

• cargo test: 162 passing (159 + 3 new)
• pytest interfaces/python/tests: 87 passing
• python -m policyengine_uk_compiled.yaml_tests tests/policy: 29/29
• End-to-end smoke: single-adult Band D household → £1,710 (vs £2,280 for two-adult), confirming the discount applies correctly

Stacking

vahid/council-tax-spd ← vahid/dla-aa-from-flags (#57) ← vahid/pip-from-flags (#56) ← vahid/lbtt-ltt (#55) ← vahid/yaml-test-harness (#54) ← vahid/parity-harness (#53) ← vahid/from-situation (#52). Seven-deep stack.

Out of scope (remaining for #42)

• Council Tax Reduction (CTB) — needs means-tested CTR scheme rules
• Status discounts (full-time students, severe mental impairment, apprentices)
• Property exemptions (Class N: all-student households, Class M, etc.)
• Devolved CTR variants (English, Scottish, Welsh schemes diverge)
• Wiring the calculated value into actual household net income (currently only the FRS-recorded value is used)

Test plan

• CI passes (cargo + pytest + parity smoke + YAML cases)
• Manual sanity-check: two-adult Band D household yields £2,280; single-adult yields £1,710
• Reform authors can override single_person_discount_rate via CouncilTaxParams

🤖 Generated with Claude Code

[vahid-ahmadi]

Rebased onto main — ready to merge.

Now independent on main and mergeable on its own. The single-person discount logic is correct — adults counted via the 18+ is_adult() predicate, 25% reduction applied only when exactly one adult is resident (LGFA 1992 s.11(1)(a); Band D £2,280 → £1,710), children correctly excluded. single_person_discount_rate defaults to 0.25 via serde, so existing param files need no changes. Verified after rebase: builds clean, all 5 council-tax unit tests pass, Python wrapper imports cleanly; YAML cases cover two-adult / single-adult / Band A / single-adult-with-children. One non-blocking note: "disregarded persons" (students etc.) aren't modelled, consistent with the rest of the codebase.