Introduce discrete state choice specific continuous grids

[MaxBlesch]

PR Summary: State-Choice-Specific Continuous Grids

Main features

1. State-choice-specific continuous grids — continuous_grid_functions, a new top-level argument to create_model_dict/setup_model, lets a grid (assets_end_of_period, assets_begin_of_period, or any additional continuous state) depend on the discrete state and choice, not just be one fixed global array. model_config["continuous_states"] entries can now be None (fully state-specific, no default array).
2. Representative-parent grid selection — since the solution lives on the state-choice space, every place that evaluates a transition into a child now distinguishes the child (drives the transition function) from a representative parent (supplies the grid values), consistency-checked once at model-build time.
3. Choice-dependence detection (transition_funcs_depend_on_choice) — when neither the budget equation nor the continuous-state transition declares choice, the law of motion is evaluated once per unique child state instead of once per state-choice, a pure cost optimization.
4. Shared batch-dedup logic — factored out of the two batch-size algorithms into child_state_dedup.py, removing duplicated code and exposing the state-level dedup that (3) relies on.
5. N-D regular-grid interpolation — new interpnd_regular.py for reading stored solutions on own (possibly state-specific) regular grids under Druedahl-Jorgensen.
6. Final period unified onto the main code path — the final period now goes through the same calc_law_of_motion entry point as every other period instead of a duplicated one.

Tests per feature

Feature	Test file(s)
(1) Config plumbing (continuous_grid_functions, None-declared grids)	test_state_specific_grids_config.py
(2) Representative-parent grid selection	test_state_specific_grids_law_of_motion.py, test_state_specific_grids_end_to_end.py
(1)+(2) end-to-end, via solve()/simulate()/choice_values_for_states	test_state_specific_grids_end_to_end.py
(1)+(2) absolute correctness (vs. closed-form and a hand-solved NumPy reference, incl. a second state-specific grid on an additional continuous state)	test_state_specific_grids_reference.py
(3) Choice-dependence granularity (bit-for-bit equivalence of the two branches)	test_state_specific_grids_law_of_motion.py
(3) Choice-dependent budget, absolute correctness over 4 periods with stochastic transitions	test_state_specific_grids_reference.py (reference.py's work_cost arg)
(4) Batch dedup consistency	test_state_specific_grids_law_of_motion.py, test_state_specific_grids_config.py
(5) N-D regular-grid interpolation	test_interpnd_regular_own_grids.py, test_interpolation.py
(6) Final period / general regression coverage	test_two_period_continuous_experience.py, test_law_of_motion.py

[codecov]

Codecov Report

❌ Patch coverage is 95.56869% with 30 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.88%. Comparing base (578ff8d) to head (a76b24d).

Files with missing lines	Patch %	Lines
src/dcegm/interfaces/combined_solve.py	81.81%	20 Missing ⚠️
src/dcegm/egm/interpolate_marginal_utility.py	78.57%	3 Missing ⚠️
src/dcegm/interfaces/sol_interface.py	33.33%	2 Missing ⚠️
src/dcegm/interpolation/interp_interfaces.py	85.71%	2 Missing ⚠️
src/dcegm/interpolation/simulation_interp.py	92.30%	1 Missing ⚠️
...ocessing/model_structure/continuous_state_grids.py	98.71%	1 Missing ⚠️
src/dcegm/solve_single_period.py	92.30%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #212      +/-   ##
==========================================
+ Coverage   93.06%   93.88%   +0.81%     
==========================================
  Files          71       74       +3     
  Lines        2597     3125     +528     
==========================================
+ Hits         2417     2934     +517     
- Misses        180      191      +11     

☔ 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.