Fix basic state pension calculation

[vahid-ahmadi]

Fixes #1090

[vahid-ahmadi]

The original issue was that the basic_state_pension formula returned 0 when simulation.dataset was None, preventing the calculation from ever reaching the amount_in_data_year logic where maximum_basic_sp would be applied. I fixed this by modifying the formula to handle the no-dataset case by using the current period for calculations instead of returning 0 early, ensuring that amount_in_data_year is always calculated and properly depends on maximum_basic_sp through the min_(reported, maximum_basic_sp) operation.

[nikhilwoodruff]

@PolicyEngine can you review this/fix issues and get enough done to get this either over the line, or closed if not worth continuing. Run individual failing tests (check CI) with policyengine-core test [path.yaml] -c policyengine_uk to avoid big memory usage and wait times

[policyengine]

I've fixed the basic state pension calculation. Here's what I changed:

Issues found in the original PR:

1. Used yoy_growth.triple_lock instead of indices.triple_lock - the former is year-over-year growth rate, the latter is the cumulative index needed for uprating
2. Non-BASIC pension types returned the reported amount instead of 0 - but basic_state_pension should only return values for BASIC type pensions
3. Test expectations were incorrect for non-BASIC types

My fix:

• Allow basic_state_pension to calculate without a dataset (the original code returned 0 immediately if no dataset)
• Keep returning 0 for non-BASIC pension types (preserves original behavior)
• Only apply triple lock uprating when using dataset data
• Use the correct indices.triple_lock cumulative index
• Fixed test expectation for non-BASIC types to expect 0

⚠️ Branch is significantly outdated:
This branch is from July 2025 and has 18 merge conflicts with master, including unrelated changes to BRMA_LHA_rate.py that cause LHA-related tests to fail (wrong case: lha vs LHA).

Recommendation:

• If the state pension fix is valuable, merge master into this branch to resolve conflicts, or
• Create a fresh PR from master with just my changes to basic_state_pension.py and the test file

[nikhilwoodruff]

Can you sort the merge conflicts @PolicyEngine

[policyengine]

Resolved all 19 merge conflicts and merged master into this PR branch.

Key resolution decisions:

• Kept the PR's fix for basic_state_pension.py that allows the calculation to work without datasets (the core fix for Check basic state pension logic #1090)
• Accepted master's refactored system.py with modular imports
• Accepted master's updated test values and uprating improvements
• Accepted all new features from master (salary sacrifice cap, student loans, etc.)

The branch is now up to date with master and should pass CI.