This fixes the invalid stripping of percent characters from JSON body inputs introduced by PR #24. Invalid UTF-8 sequences in the JSON itself are still sanitized, but the data being represented by JSON is not necessarily UTF-8 sequences and so performing URI decoding of percent sequences is inappropriate.
Thanks for spotting this bug, @cosine! I wanted to make a few changes to your implementation, so I'm closing this in favor of #27. I'll get a new release out shortly.
This fixes the invalid stripping of percent characters from JSON body inputs introduced by PR #24. Invalid UTF-8 sequences in the JSON itself are still sanitized, but the data being represented by JSON is not necessarily UTF-8 sequences and so performing URI decoding of percent sequences is inappropriate.