Fix UnicodeEncodeError on Windows (cp1252 default)

[cnrdfrncs-droid]

Summary

scripts/watch.py crashes on Windows when it prints common Unicode characters (the arrow U+2192, em-dashes, ellipses) to stdout/stderr. Both streams default to cp1252 on Windows, which doesn't include those code points.

Repro

On a fresh Windows install with Python 3.13:

python scripts/watch.py "https://www.youtube.com/watch?v=jNQXAC9IVRw" --start 0 --end 10 --no-whisper

Crashes at the focus-range header line:

UnicodeEncodeError: 'charmap' codec can't encode character '→' in position 25: character maps to <undefined>
  File "scripts\watch.py", line 158, in main
    print(f"- **Focus range:** {format_time(...)} → {format_time(...)} ...")

The pipeline itself (download, captions, frame extraction) all succeeds — only the final markdown report fails to render.

Fix

Force UTF-8 on sys.stdout / sys.stderr at script startup via .reconfigure(encoding="utf-8"). The hasattr guard makes it a no-op on streams without that method (e.g. when stdout is replaced by something non-standard). On macOS/Linux this is also a no-op since both streams are already UTF-8.

Alternatives considered:

• Setting PYTHONUTF8=1 env var — works but pushes the burden onto every user. Not discoverable.
• Replacing every Unicode char in the source (→ → ->, em-dash → --) — invasive, easy to miss occurrences, and whisper.py / other scripts have the same chars.
• Wrapping sys.stdout with a TextIOWrapper — equivalent but more code.

reconfigure is the smallest, most contained fix.

Test plan

• Verified on Windows 11 / Python 3.13 / cp1252 console — python scripts/watch.py <url> --start 0 --end 10 --no-whisper runs to completion without PYTHONUTF8 set, prints the → character cleanly in the focus-range header and final report.
• No behavior change on macOS/Linux (streams already UTF-8).