fix: use UTF-8 for text file operations

[Bill-Billion]

Summary

• make course, reference agent, and Agent Builder text I/O explicitly use UTF-8
• cover file tools, skill manifests, task state, transcripts, and team inboxes without adding encoding detection dependencies
• sync multilingual code examples and regenerate tracked Web snapshots
• add non-ASCII behavior tests and a regression guard for implicit text encodings

Verification

• python -m pytest (279 passed)
• full suite with LC_ALL=C, PYTHONUTF8=0, and locale coercion disabled (279 passed)
• cd web && npm run build
• git diff --check

Closes #542