Reference creatortrack dev base utf8 rebuild

The creatortrack_dev_base template database was originally created as SQL_ASCII (inherited from template1 defaults) instead of UTF8. This caused psycopg to return text columns as bytes objects, breaking migration tracking (every migration looked pending, causing duplicate-key errors on INSERT).

Fix applied 2026-06-20: - forge_suite_dev_base.sh now explicitly creates the base DB as ENCODING='UTF8' LC_COLLATE='C.UTF-8' LC_CTYPE='C.UTF-8' TEMPLATE=template0 - Builder also auto-baselines existing suite migrations so clones never re-run already-baked schema - forge_workspace_migrate.py got a defensive bytes-decode for belt-and-suspenders

Production lifeos DB is UTF8/C.UTF-8 — clones must match or unicode corruption risk exists for emoji/accented content.

How to apply: If a future dev base rebuild is needed, always pass explicit encoding + template0. Never rely on template1 defaults. If migration runner reports everything as pending despite applied migrations, check DB encoding first (SELECT pg_encoding_to_char(encoding) FROM pg_database WHERE datname = current_database();).

[auto-memory session 92503a43-880c-4800-b081-18fc1658f0e6, confidence 0.85, mode direct]