Add discrete per-filing-status AGI schedule to MD/OH dependent exemption contrib

[DTrim99]

Resolves #8799.

Problem

The dashboard drives most states' dependent exemption through the gov.contrib.states.{st}.dependent_exemption contrib (us#8696). That contrib offers a flat amount (or -1 sentinel reading baseline), a single linear phase-out, and age_limit — which fits RI cleanly. MD and OH don't fit, because their dependent exemption is a discrete, income-stepped, per-filing-status schedule, so consumers had to reach into baseline brackets instead.

Change

Give MD and OH a first-class schedule sub-tree in their contrib, mirroring the baseline, so the discrete step-down is adjustable entirely through the contrib (aligned with the rest of the us#8696 family):

• MD — per-filing-status bracket params schedule/{single,separate,joint,head,surviving_spouse}.yaml, mirroring the baseline §10-211 AGI step-down ($3,200 → $1,600 → $800 → $0 with filing-status-specific cutoffs).
• OH — schedule/amount.yaml, a MAGI-stepped bracket param mirroring the baseline §5747.025 schedule ($2,400 → $2,150 → $1,900 → $0).

The reform's dependent-exemption maximum now reads the stepped per-dependent amount from the contrib schedule rather than the baseline params. Defaults equal baseline, so:

• the -1 amount sentinel remains a no-op at default (verified by the existing no-op tests),
• a non-negative flat amount still overrides the schedule,
• the linear phaseout still applies on top.

Tests

Added to each state's contrib reform test file:

• default schedule reproduces the AGI/MAGI step-down — a higher-income unit lands on a lower step purely through the contrib (MD single AGI 130k → $800/exemption; OH MAGI 50k → $2,150/exemption).
• schedule per-dependent amount is adjustable — overriding schedule.single[0].amount (MD) / schedule.amount[0].amount (OH) changes only the dependent portion, leaving the baseline personal portion intact.

All existing no-op / flat-amount / phase-out / age-limit cases are unchanged.

🤖 Generated with Claude Code

[daphnehanse11]

PolicyEngine Review: PR #8800

Recommendation

COMMENT / rerun Quick Feedback before merge.

I found no code, source-value, reference, or test-coverage defects in the PR itself. The only blocking status is the red Quick Feedback check, whose log shows a GitHub runner shutdown/cancellation rather than a test assertion failure. Full suite, lint, changelog, and metadata checks passed.

Findings

No code findings requiring changes.

CI Note

• Quick Feedback (Selective Tests + Coverage) is red.
• Log evidence: selective tests started, collected 26 items, printed Maryland pass dots and Ohio pass dots, then reported: runner received a shutdown signal / operation was canceled.
• Local targeted run of the two changed contrib test files reached: 26 passed, 1 warning in 215.09s (0:03:35).
• Required action before merge: rerun Quick Feedback so the PR has a green required check.

Scope Reviewed

PR #8800 adds first-class contrib schedules for Maryland and Ohio dependent exemption reforms:

• Maryland: gov.contrib.states.md.dependent_exemption.schedule.{single,separate,joint,head,surviving_spouse}.
• Ohio: gov.contrib.states.oh.dependent_exemption.schedule.amount.
• Reform logic now uses those contrib schedules when amount remains the negative sentinel, while flat non-negative amount overrides still work.
• Tests add default schedule and schedule override scenarios for both states.

Fresh worktree: /tmp/policyengine-us-pr-8800-review at 68f292dc8bb6e98897c2affbc8dda53268db29ec.

Regulatory And Source Review

Maryland:

• Official 2025 Resident Instruction Booklet, file page 12, Exemption Amount Chart (10A), supports:
  • single/separate: $3,200 up to $100,000; $1,600 over $100,000 through $125,000; $800 over $125,000 through $150,000; $0 above $150,000.
  • joint/head/surviving spouse: $3,200 up to $150,000; $1,600 over $150,000 through $175,000; $800 over $175,000 through $200,000; $0 above $200,000.
• PR Maryland schedules match these values and use right=True, consistent with baseline md_personal_exemption and the source's "over / but not over" boundaries.

Ohio:

• Official 2025 IT 1040 booklet, file page 17, supports:
  • $40,000 or less -> $2,400;
  • $40,001-$80,000 -> $2,150;
  • $80,001-$749,999 -> $1,900;
  • $750,000 or greater -> $0.
• PR Ohio schedule matches the baseline schedule and the 2025 booklet. The 2026 $500,000 cutoff follows Ohio Rev. Code 5747.025.

PDF artifacts:

• /tmp/pr-8800-review-pdf-md.pdf
• /tmp/pr-8800-review-pdf-md-page12.txt
• /tmp/pr-8800-review-pdf-md-page12-layout.txt
• /tmp/pr-8800-review-pdf-md-page12.png
• /tmp/pr-8800-review-pdf-oh.pdf
• /tmp/pr-8800-review-pdf-oh-page17.txt
• /tmp/pr-8800-review-pdf-oh-page17-layout.txt
• /tmp/pr-8800-review-pdf-oh-page17.png

Code Review

Maryland:

• The new filing-status dispatch mirrors baseline md_personal_exemption.
• adjusted_gross_income is the same AGI measure used by baseline Maryland schedule logic.
• Legal values are parameterized; no new hard-coded legal amounts in formulas.
• The negative amount sentinel remains a no-op by default because the contrib schedule mirrors baseline.

Ohio:

• oh_dependent_exemption_maximum uses p.schedule.amount.calc(agi), matching baseline Ohio schedule behavior.
• oh_agi is consistent with baseline oh_personal_exemptions.
• Flat amount override behavior is preserved.

No vectorization, aggregation, period, or code-organization defect found.

Test Review

Changed tests cover:

• default schedule reproducing the AGI/MAGI step-down for Maryland and Ohio;
• schedule override changing only the dependent portion;
• existing no-op, flat override, phaseout, age-limit, and downstream-flow cases remain.

Local command:

uv run python -m policyengine_core.scripts.policyengine_command test -c policyengine_us policyengine_us/tests/policy/contrib/states/md/dependent_exemption_reform_test.yaml policyengine_us/tests/policy/contrib/states/oh/dependent_exemption_reform_test.yaml

Result: 26 passed, 1 warning in 215.09s (0:03:35). The process then hung in pytest cleanup (gc.collect()); I interrupted after the pass summary was emitted.

Report Files

• /tmp/pr-8800-review-context.md
• /tmp/pr-8800-review-pdf-manifest.md
• /tmp/pr-8800-review-regulatory.md
• /tmp/pr-8800-review-references.md
• /tmp/pr-8800-review-code.md
• /tmp/pr-8800-review-tests.md
• /tmp/pr-8800-review-pdf-schedules.md
• /tmp/pr-8800-review-pages.md