Review follow-up for the D1 charset benchmark.
- XMLTV: feed the parser 64 Ki-character slices of one decoded string.
Decoding each 64 KiB buffer separately left the BOM control one-byte
after its first chunk and could split Cyrillic characters into U+FFFD.
- Verify parsed M3U names/groups and XMLTV titles against the fixture in
an untimed pass, including that no title contains U+FFFD.
- Rotate the starting input every round so no variant always runs first.
- Move workload definitions to charset-parse-workloads.ts to keep the
runner under the line target.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
