fix: write UTF-8 bytes directly to stdout to avoid encoding errors

[fuleinist]

Fix: stdout Unicode encoding errors on Windows

On Windows (and other platforms) where sys.stdout.encoding is limited (e.g., cp1252, gbk), piping markitdown output to a file causes UnicodeEncodeError for characters outside the target encoding.

Problem

The previous workaround of encode+decode with errors='replace' still failed when stdout.encoding was None (piped stdout on some platforms), and didn't solve the root issue of stdout's limited text encoding.

Solution

Write UTF-8 bytes directly to sys.stdout.buffer, which:

• Bypasses stdout's text encoding limitation
• Works reliably when stdout is piped or redirected
• Handles all Unicode characters correctly
• Falls back to print() with encoding='utf-8' for unusual cases

Issues Fixed

• Fixes UnicodeEncodeError #1788 (UnicodeEncodeError on Windows GBK)
• Fixes Unicode error converting Microsoft Documentation PDF to markdown file. #1597 (UnicodeEncodeError on Windows cp1252)

Testing

• Verified CLI works with Unicode input via stdin
• Import test passes
• CLI version check passes