Xia Li-yao authored May 01, 2021 and Bodigrim committed May 22, 2021

At the beginning of a new chunk we may be trying to complete a UTF-8
sequence started in the previous chunk (contained in the `undecode0`
buffer). If it turns out to be invalid, we must apply the `onErr`
handler to every character in that buffer.

When we reach the end of the chunk, we must also be more careful
about when to keep the previous buffer: a UTF-8 sequence (up to 4 bytes)
can span more than two chunks, when those chunks are very short
(of length 0, 1, or 2).