feat: add build_period axis to investment effect matrices

[FBumann]

Summary

Investment effect fields (effects_per_size, effects_fixed, effects_per_size_periodic, effects_fixed_periodic) currently support an optional period dimension (added in #94). Per the design decision in #88 and the target in #92, they should also support a build_period axis to enable period-dependent construction costs.

Target dimensions

Field	Current	Target
Investment.effects_per_size	(invest_flow, effect, period?)	(invest_flow, effect, period?, build_period?)
Investment.effects_fixed	(invest_flow, effect, period?)	(invest_flow, effect, period?, build_period?)
Investment.effects_per_size_periodic	(invest_flow, effect, period?)	(invest_flow, effect, period?, build_period?)
Investment.effects_fixed_periodic	(invest_flow, effect, period?)	(invest_flow, effect, period?, build_period?)

Construction rule (from design decision)

• Scalar → constant across all periods → diagonal
• 1D (build_period,) → diagonal (cost charged in the period when built)
• 2D (period, build_period) → as-is (construction spread, learning curves, decommissioning costs)

Diagonal means: effects[p, bp] = value[bp] only where p == bp.

Scope

• model_data.py: _InvestmentArrays.build must accept and expand build_period-indexed inputs via the diagonal rule
• model.py: consumption multiplies invest_size_at_build[flow, build_period] by the 2D effect matrix and sums over build_period
• elements.py: type hints already accept TimeSeries (which includes xr.DataArray), so no change needed
• Tests: add cases for scalar, 1D (build_period,), and 2D (period, build_period) inputs

Context

Separated from #92 (period-aware effects) and #94 (implementation). The _EffectTemplate helper introduced in #95 can be extended to support this.