LibTextCodec: Use String::from_utf8() when decoding UTF-8 to UTF-8
This way, we still perform UTF-8 validation, but don't go through the
slow generic code path that rebuilds the decoded string one code point
at a time.
This was a bottleneck when loading a canned copy of reddit.com, which
ended up being ~120 MiB large.
- Time spent decoding UTF-8 before this change: 1192 ms
- Time spent decoding UTF-8 after this change: 154 ms
That's still a long time, but 7.7x faster is nothing to sneeze at! :^)
Note that if the input fails UTF-8 validation, we still fall back to
the slow path and insert replacement characters per the WHATWG Encoding
spec: https://encoding.spec.whatwg.org/#utf-8-decode