fix(tee): prevent panic on UTF-8 multi-byte truncation boundary

[pszymkowiak]

Summary

• &raw[..max_file_size] in write_tee_file panics if the byte offset falls inside a multi-byte UTF-8 character (Japanese, emoji, etc.)
• Now uses char_indices() to find the nearest safe boundary before slicing
• 2 new tests: Japanese 3-byte chars + emoji 4-byte chars at truncation boundary

Ref: #640 (L-3 finding from security audit)

Test plan

• 1360 tests pass, 6 ignored
• test_write_tee_file_truncation_utf8_boundary — 3-byte Japanese chars, cut at 998/999
• test_write_tee_file_truncation_emoji — 4-byte emoji, cut at 201/400
• Python proof: old code would decode-error at byte 996-997