State case-bucket terms ignore the bucket years in the targets #5

The 5-year case terms sum fixed simulated years and never read the bucket years in the targets (R/IMIS_functions_st.r:99-153, R/calib_functions_st.r:8-9). Not introduced in this fork.

• PPML/MITUS fits states from the 10-29-21 file, whose buckets match the summed years
• this fork loads one targets file per location (R/data_loadmodel.R:60), so states fit the 04-20-22 buckets one year out of step
• county targets built in the 04-20-22 layout are out of step the same way, and their tables can end in different years
• published state calibrations do not reproduce on this fork

Bucket labels in the PPML/MITUS targets contain errors that would block reading years from them:

• X2015.1019 for 2015-2019 in cases_nat_st_5yr, 10-29-21
• 99_13 for 1999-2003 in rt_fb_cases_sm, both files
• 00_04 for 2001-2005 and 06_10 twice in hr_cases_sm, 04-20-22

Changes

• each bucket term reads its bucket years from the targets and sums the simulation over those years
• model_load stops with an error naming the table when a bucketed table is not contiguous or its last bucket does not end in last_weight_year
• the targets tables keep their current formats, and bucketing yearly data is left to each location's data build

Fix on branch generalize-calibration-files.

Line numbers refer to branch generalize-calibration-files at d06b730b, as of 2026-11-05.