Add memoization cache to get_id() with signal-based invalidation

[cmutel]

Summary

• Adds a module-level dict cache to get_id() for (database, code) → id lookups, avoiding repeated SQLite queries during large imports (reported to save ~1 minute on ecoinvent imports in Adding memoization (cache) to 'get_id()', saves 1 minute per ecoinvent import. #254)
• Cache is invalidated via blinker signal handlers on all relevant events: project switch, database delete/reset/write, individual activity deletion, and activity code or database changes
• Chose a plain dict over functools.lru_cache because selective per-database and per-activity invalidation requires __delitem__, which lru_cache does not expose

Test plan

• test_get_id_cache_populates — cache is written on first call
• test_get_id_cache_returns_same_value — cache hit returns correct value
• test_get_id_cache_project_changed — project_changed clears entire cache
• test_get_id_cache_database_delete — on_database_delete removes only that database's entries
• test_get_id_cache_database_reset — on_database_reset removes only that database's entries
• test_get_id_cache_database_write — on_database_write removes only that database's entries
• test_get_id_cache_activity_delete — signaleddataset_on_delete removes only that activity's entry
• test_get_id_cache_activity_code_change — on_activity_code_change removes old key, leaves others
• test_get_id_cache_activity_database_change — on_activity_database_change removes old key, leaves others

Closes #254