Add a `Utf8CharDecoder` type for incrementally decoding UTF-8 byte-by-byte

[jwodder]

#[derive(Clone, Debug, Eq, PartialEq)]
struct Utf8CharDecoder(...);

impl Utf8CharDecoder {
    fn new() -> Self { todo!() }
    fn feed(&mut self, byte: u8) -> Option<Result<char, SomeError>> { todo!() }
    fn finish(self) -> Result<(), SomeError> { todo!() }
    fn reset(&mut self) { todo!() }
    // Something for decomposing into the inner partial bytes
    // Something for testing whether there are currently partial bytes stored
    // Something for querying how many continuation bytes are needed to complete the current character
}

• Also add a lossy variant

• Error conditions that the error type must cover:

  • Codepoint is greater than 0x10FFFF
  • Codepoint is a surrogate character
  • Non-canonical encoding of codepoint (e.g., 0b1100_0000 0b1000_0000 for the NUL byte)
  • Partial UTF-8 sequence followed by non-continuation byte (i.e., ASCII char or new start byte)
  • Continuation byte encountered without preceding matching start byte

• On an error, reset the decoder to the initial state?