Fix the Defined Benefits and Points System pensions so they can actually run

[marcelolafleur]

The non-US pension systems in pensions.py crash when used with a real Specifications object. We found this calibrating OG-BRA: Brazil's pension system (RGPS) is a defined benefits scheme, so OG-BRA is the first country model to use pension_system = "Defined Benefits" — and it crashes on the first solve. Same family of bugs as the Points System crash reported in #1075.

The cause is that these code paths were never updated after some parameters became time-varying: the functions pass arrays where the numba loops expect scalars, and scalars where they expect arrays. The existing tests didn't catch this because they feed the loops pre-cleaned inputs from mock objects instead of a real Specifications object.

Three changes (the third was added after testing a full transition path downstream):

• Fix the Defined Benefits and Points System paths so they run. They use the steady-state retirement age and earnings profile for now; full time variation stays open in Defined Benefits pension system not handling parameters object sizes correctly #1014. The NDC system needs more than this — its growth-rate settings aren't in default_parameters.json yet — so it remains unusable for now.
• Add tests that call the pension functions with a real Specifications object (these fail on master with the Bug: Points System crashes with ValueError due to p.retire being a list (plus subsequent TypingError) #1075 errors), plus a steady-state solve test with the DB system.
• Make the DB system solve the transition path, not just the steady state. Two pieces were missing: the benefit formulas reference pre-time-path wages and labor that nothing ever supplied (labor now comes from the model's initial condition — the same baseline object that initializes wealth — and wages are anchored to the period-0 wage of the current path), and there was no evaluation of DB amounts over the full (T, S, J) path for the TPI's aggregate revenue step (added as DB_3dim_loop, built from each cohort's own wage history so aggregates match household behavior; it reproduces the steady-state solution to machine precision on a constant path).

Fixes #1075. Part of #1014.

Tested: full pension and tax suites pass (71 tests); the DB steady state solves. Downstream check: the OG-BRA Defined Benefits calibration solves both the baseline and a reform transition path against this branch, passes the resource-constraint check on both, and hits its pension spending target to four decimals.

cc @rickecon @jdebacker

[codecov-commenter]

Codecov Report

❌ Patch coverage is 58.00000% with 21 lines in your changes missing coverage. Please review.
✅ Project coverage is 72.92%. Comparing base (7cae932) to head (6062e0a).
⚠️ Report is 2 commits behind head on master.

Files with missing lines	Patch %	Lines
ogcore/pensions.py	56.25%	21 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1167      +/-   ##
==========================================
- Coverage   72.95%   72.92%   -0.04%     
==========================================
  Files          22       22              
  Lines        5480     5525      +45     
==========================================
+ Hits         3998     4029      +31     
- Misses       1482     1496      +14     

Flag	Coverage Δ
unittests	72.92% <58.00%> (-0.04%)	⬇️

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

Files with missing lines	Coverage Δ
ogcore/TPI.py	36.29% <100.00%> (+0.11%)	⬆️
ogcore/__init__.py	100.00% <100.00%> (ø)
ogcore/pensions.py	67.30% <56.25%> (-0.93%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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

[jdebacker]

@marcelolafleur I appreciate the clarity in this PR about what is addressed and what is not. While not totally solving all of Issue #1014, I think this PR is an important addition and fix of the DB and PS pension systems. I want to merge this and create a new release. Can you update the version to 0.17.0 and include in the change log not just this PR, but the changes in PRs #1174, #1171, and #1166?

[marcelolafleur]

Thanks @jdebacker. All set:

Version bumped to 0.17.0 (pyproject.toml and ogcore/init.py).
New [0.17.0] changelog section covering this PR plus #1174, #1171, and #1166.
Also corrected a stale docstring in DB_3dim_loop that said the function wasn't jitted — it is in the current version of numba.

Ready to merge whenever you are.