Snt25 659 incidence

[EstebanMontandon]

Summary

Migrates snt_dhis2_incidence's computation script (snt_dhis2_incidence.ipynb + snt_dhis2_incidence.r) off its bespoke setup helpers onto the shared init_snt_workspace() / load_snt_config() / load_dataset_file() pattern, and fixes two latent bugs surfaced along the way.

• Setup consolidation: replaced load_utils() / setup_paths() / create_intermediate_data_dir() / install_and_load() / set_env_openhexa() / import_config_json() / config_generic() / config_incidence() / set_fixed_cols() with init_snt_workspace() + load_snt_config() (from code/snt_utils.r) and equivalent inline config assignments. Output paths (INTERMEDIATE_DATA_PATH, DATA_PATH) are unchanged in practice.
• Fix: careseeking file loading could crash the whole run. read_csv(CARESEEKING_FILE_PATH, ...) is now wrapped in tryCatch — a corrupted/malformed user-provided file now falls back to DHS data instead of stopping the pipeline. Also simplified the two-flag IMPORTED_CARESEEKING_FROM_FILE / IMPORTED_CARESEEKING_FROM_DHS tracking down to one flag, since "not from file" now unambiguously means "from DHS."
• Fix: join_careseeking_data() silently continued on bad input. When neither ADM1_ID nor ADM2_ID is present, it now stop()s with a clear message instead of logging an "error" severity and falling through to a confusing downstream crash.
• Reduced implicit global-variable coupling between functions: check_fixed_cols_in_routine(), check_dhis2_indicators_cols_in_routine(), check_pres_col() (renamed from check_PRES_col), load_population_data(), select_routine_dataset_and_filename(), and prepare_disaggregated_indicators()'s call site now take/return explicit values instead of reading or writing globals implicitly.
• Removed dead code: build_monthly_cases() / build_yearly_incidence() were unused (the notebook has always used inline equivalent logic) and had already drifted from it.
• Typo fix: coherence_checkes_yearly_incidence() → coherence_check_yearly_incidence().

Follow-up (not fixed in this PR)

snt_dhis2_incidence.r still has ~18 functions that set globals via <<- instead of returning a value (prepare_disaggregated_indicators(), select_population_column(), load_dhs_careseeking_data(), load_reporting_rate_data(), check_reporting_rate_data(), enforce_numeric_cols(), handle_zeros_in_reporting_rate(), and the coherence_check_*() family), which snt_dhis2_incidence.ipynb then reads back out of the global environment.

We should try to avoid functions that set global variables the notebook reads back later in an obscure way , at the end this tends to hide side effects behind what looks like a simple call, and makes debugging harder since the cause of a bug can be several cells away from where it surfaces (missing references to variables secretly created). We should also aim for generic, reusable functions instead of wrapping notebook cell computations.