fix(io): tool_output falls back to ASCII on UnicodeEncodeError (#5029)

[MukundaKatta]

Why

Closes #5029.

/map output on Windows with cp1251 (Russian) locale crashed with an uncaught UnicodeEncodeError. Root cause: Rich uses U+22EE (vertical ellipsis, ⋮) for tree/box drawing, and the legacy Windows terminal's cp1251 codec can't encode it.

The crash traceback points at aider/io.py:1012 inside tool_output:

File "io.py", line 1012, in tool_output
    self.console.print(*messages, style=style)

What

_tool_message already handles this exact case (see current code lines 979–986): catch UnicodeEncodeError, re-encode with errors='replace', retry the print.

tool_output was missing the same handler. This PR mirrors the existing fallback into tool_output — same pattern, same semantics. No new public API, no behavior change on UTF-8 platforms.

Diff shape

• aider/io.py: +14 lines (try/except in tool_output)
• tests/basic/test_io.py: +26 lines (regression test)

Test

New test test_tool_output_unicode_fallback models the exact #5029 scenario — U+22EE message, first console.print raises UnicodeEncodeError, second succeeds with the character replaced.

$ pytest tests/basic/test_io.py -k unicode -v
test_autocompleter_with_unicode_file PASSED
test_tool_message_unicode_fallback PASSED
test_tool_output_unicode_fallback PASSED
3 passed

Scope

Deliberately narrow — just the crash in #5029. Broader robustness (forcing UTF-8 console mode on Windows, warning users, etc.) is a larger design discussion and not needed to close this issue.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}