fix(cicd): chunk bot messages on character boundaries

[anxkhn]

GithubController._chunk_up_api_message splits a message into chunks that fit
GitHub's per-field byte limit (MAX_BYTE_LENGTH, 65535). It did this by UTF-8
encoding the whole message, slicing the resulting bytes at fixed
MAX_BYTE_LENGTH offsets, and decoding each slice with errors="ignore":

message_encoded = message.encode("utf-8")
return [
    message_encoded[i : i + self.MAX_BYTE_LENGTH].decode("utf-8", "ignore")
    for i in range(0, len(message_encoded), self.MAX_BYTE_LENGTH)
]

When a multibyte character (accented letter, CJK, emoji, etc.) straddles a chunk
boundary, its UTF-8 bytes are split across two slices. decode("utf-8", "ignore")
then discards the incomplete byte sequence on both sides, so that character is
silently dropped from the bot's PR comment or check-run output.

This change accumulates characters into a chunk while the running UTF-8 byte
length stays within MAX_BYTE_LENGTH, so a boundary is only ever placed between
whole characters and no character is bisected. Behavior for the common case is
unchanged: ASCII-only messages chunk into the exact same pieces as before, and an
empty message still yields no chunks. Both existing callers rely only on the
number of chunks (comment edit unpacks body, *truncated; check-run output
unpacks summary, text, *truncated), and that contract is preserved.

Test Plan

• Added test_chunk_up_api_message_preserves_multibyte_chars, which builds a
  message just over MAX_BYTE_LENGTH with a multibyte character (é) placed on
  the byte boundary between two chunks, and asserts the chunks reassemble into the
  original message and each chunk stays within MAX_BYTE_LENGTH bytes. This test
  fails on the previous byte-slicing implementation (the boundary character is
  dropped) and passes with the fix.
• pytest tests/integrations/github/cicd/ -m "not slow and not docker" -> 55
  passed, 28 deselected.
• make style (ruff check + ruff format) clean on the changed files.

Checklist

• I have run make style and fixed any issues
• I have added tests for my changes (if applicable)
• All existing tests pass (make fast-test)
• My commits are signed off (git commit -s) per the DCO