Also minor refactoring in decoding in order to tolerate the absence of a charset and treat as US_ASCII. See gh-23485