Require NOT NULL iteration and stage keys in opchar input tables

[anamileva]

Summary

The iteration-keyed opchar input tables (variable generator profiles, hydro operational chars, energy profiles, horizon shaping, load modifier profiles, load component shift bounds, exogenous state of charge, variable OM by timepoint) are read with equality filters on weather_iteration, hydro_iteration, and stage_id (e.g. weather_iteration = 0 for inputs that do not vary by weather iteration). NULL never satisfies an equality filter, so a blank key cell in an input CSV silently dropped the row: the project ended up with no data and the scenario failed at model load rather than at import or validation.

• Declare these 24 columns NOT NULL in db/db_schema.sql, so a blank cell now fails at CSV import with NOT NULL constraint failed: <table>.<column>.
• Add tests/test_db_schema_iteration_keys.py, which discovers the iteration-keyed tables from the schema (every inputs_project_* table with a companion _iterations table keyed by an opchar column) and asserts the constraint on each key column, so a future table of this kind cannot be added without it.

Notes

• Databases created before this change keep their unconstrained tables until rebuilt.
• No committed example CSV has a blank key cell (scanned all 188 key columns under db/csvs_test_examples); the examples database rebuilds from CSVs under the new schema and the example scenarios touching these tables pass.
• The test was mutation-checked: stripping one constraint from the schema makes it fail.

🤖 Generated with Claude Code