There's a number of changes in here:
- Plausibility checks to avoid expensive decoding are now done everywhere,
  speeding up usage from JS.
- Both positive and negative decoding results are cached, avoiding double
  decoding when we have to do multiple passes (e.g. generic and custom).
- With plausibility checks in the backend, JS scripts don't have to deal
  with that anymore.
- Binary or textual content is now auto-detected.
- Checking for multiple code types at once is not supported, which allows
  to reuse the binarization computation.

Overall this produces a speedup of about 15% on a double-pass PDF ticket.