Add WPP-2024 reference series (median + 95% PI) to population_by_age_and_sex, population_by_broad_age_group, annual_growth_rate

[cimentadaj]

Background

run_forecast() already returns WPP-2024 reference columns (un_*_median / un_*_95low / un_*_95high) for the time-series outputs:

• population_by_time
• tfr_by_time
• e0_by_time
• mig_by_time

This lets the demographydash app overlay the UN reference on those charts. The remaining by-age outputs do not carry UN reference columns:

• population_by_age_and_sex — has only year, age, popM, popF
• population_by_broad_age_group — has only year, age, pop, pop_percent
• annual_growth_rate — has only year, age, growth_rate

So the demographydash Compare-Simulations views for population pyramid, population by broad age groups, and growth rate by age cannot draw the UN reference. This was raised by Mengjia Liang / Patrick Gerland in the April 8 OPPP review (item 5 of their feedback list) and is the only deferred item from that round.

Desired changes (mirroring the existing un_* naming in population_by_time)

1. population_by_age_and_sex — add un_popM_median, un_popF_median (and ideally un_popM_95low / un_popM_95high / un_popF_95low / un_popF_95high).
2. population_by_broad_age_group — add un_pop_median (and un_pop_95low / un_pop_95high).
3. annual_growth_rate — add un_growth_rate_median (and un_growth_rate_95low / un_growth_rate_95high).

The canonical pattern to copy is get_pop_by_time() in R/forecast.R, which already merges WPP via get_wpp_pop_by_age_multiple_years() for the median and get_wpp_indicator_multiple_years("popprojAgeGrp1dt", package = "wpp2024extra") for the prediction intervals.

Draft PR (median-only)

I'm opening a draft PR (#) that wires up the median columns for all three outputs by reusing the existing helpers get_wpp_pop_by_age_multiple_years(), get_pop_by_broad_age(), and get_annual_growth_rate(). The 95% PI columns are marked TODO(hana) — they depend on bayesPop's quantiles{M,F}age object (already touched in extract_pop_by_age_sex), and figuring out the right quantile keys / open-age handling is your call. Happy to extend the PR once you point me at the right shape.

cc @hanase