Fix UnicodeEncodeError on Windows when video titles contain emoji

[drlee91]

Problem

On Windows, Python's stdout/stderr default to the locale code page (typically cp1252). When a video title contains emoji or any character outside that code page, printing the report raises UnicodeEncodeError and watch.py exits 1 — after the download, frame extraction, and transcription have already succeeded. The frame paths are never printed, so the whole run is lost.

Emoji in titles are common on YouTube, so on a default Windows setup this breaks a large share of videos.

Repro (Windows, e.g. German locale / cp1252)

python scripts/watch.py https://www.youtube.com/watch?v=pl3n9o_ZR9M

File "...\Lib\encodings\cp1252.py", line 19, in encode
    return codecs.charmap_encode(input,self.errors,encoding_table)[0]
UnicodeEncodeError: 'charmap' codec can't encode character '\U0001f92f' in position 41: character maps to <undefined>

(The title is Anfängerin überrascht ALLE! 🤯 Angeln lernen von A bis Z 🎣 — the 🤯 kills the report.)

Fix

Reconfigure sys.stdout/sys.stderr to UTF-8 with errors="replace" at startup of the entry point. This is a no-op on macOS/Linux (streams are UTF-8 already) and is guarded with try/except for exotic streams that don't support reconfigure().

Since watch.py imports all other modules into the same process, this covers the whole pipeline.

Verification

On Windows 11 (cp1252 console):

• Before: the repro above crashes with exit 1.
• After: the same command prints the full report including the emoji title, exit 0.
• python -c "print('🤯')" still crashes in the same shell, confirming the environment itself reproduces the bug and the fix is what resolves it.

Complements #4, which fixed the same class of issue for config file I/O.

[RomeTheDev-AT]

Independent confirmation on a separate machine — Windows 11, German locale
(cp1252), Python 3.12.10. The fix is correct and resolves the crash for me.

One addition worth noting: it covers a second trigger beyond emoji titles.
Focused mode (--start/--end) prints a → (U+2192) in the "Focus range"
line of the report, which also isn't in cp1252 and crashes identically before
this patch:

python scripts/watch.py video.mp4 --start 0:30 --end 0:36
...
UnicodeEncodeError: 'charmap' codec can't encode character '\u2192'
in position 25: character maps to <undefined>

After reconfigure(encoding="utf-8", errors="replace") the full report prints
and exits 0 — for both emoji titles and the → range line. So the fix
generalizes to every non-cp1252 glyph in the report, not just titles. +1 to merge.