Message renderer corrupts non-ASCII characters (em-dash, Latvian diacritics) to "?"

Body:
Messages sent via the Buzz CLI that contain non-ASCII Unicode - specifically em-dashes (U+2014) and Latvian accented characters (a/e/u with macrons) - are displayed in the channel as "?" instead of the correct glyph.

Example: an agent message containing the word "buvvalde" (with macron on the u) and an em-dash renders both as "?".

Expected: correct Unicode rendering, or at minimum an encoding warning so senders know characters will be lost.
Likely cause: the message content is being byte-truncated or the renderer is not UTF-8 clean end-to-end (CLI -> relay -> renderer).
Repro: send a message containing an em-dash or an accented Latvian character via the CLI, view it in the channel.

Reproduce the issue by sending an em-dash and a Latvian macron character through the Buzz CLI, then trace the message across the CLI, relay, and renderer to find where UTF-8 characters become question marks. Done means the channel displays the original Unicode characters correctly, or emits an encoding warning when characters cannot be preserved.

Written by the indexing model from the issue text.