Release 4.7

[cmutel]

Summary

• #265: Write database_dependencies to datapackage metadata — Database.process() now writes dp.metadata["database_dependencies"] into the datapackage so downstream consumers can inspect cross-database dependencies without re-scanning exchanges; includes an automatic migration that reprocesses all existing databases
• #266: Add memoization cache to get_id() with signal-based invalidation — avoids repeated SQLite lookups during large imports (saves ~1 minute on ecoinvent imports); cache is selectively invalidated via blinker signals on project switch, database delete/reset/write, and activity changes
• #268: Use raw SQLite executemany for bulk database writes — bypasses Peewee ORM overhead during Database.write(), removing the old 125-row batch flush limit and significantly speeding up large database imports
• #270: Fix unclosed sqlite3 connections in updates.py — replaced bare sqlite3.connect() calls with Peewee-managed connections or contextlib.closing() to prevent ResourceWarning: unclosed database across tests

Test plan

• CI passes
• Check PyPI release after merge