String encoding/decoding round trip is wrong for multibyte UTF-8 codepoints

[huonw]

Strings are encoded with a prefix of their byte length, but the decoder takes that length and reads that many chars (aka unicode codepoints/Unicode scalar values). E.g. å is 2 bytes in UTF-8 so encoding "å" will write [2, 0xXX, 0xYY] (I don't know the exact encoding of å), and then decoding will first read the 2 bytes of the å and then try to read another codepoint from the stream, even though there's no string data left.

[TyOverby]

Fixed in 79bb788

@huonw: thanks for the bug report!