`utf-7-imap` incremental decoder and stream reader fail on split input

utf-7-imap registers an incremental decoder and a stream reader that call the one-shot decoder with no state, so a shift sequence split across chunks raises instead of being buffered. io.TextIOWrapper.read(n), codecs.iterdecode and codecs.StreamReader.read all fail on bytes the codec's own encoder produced.

import io

text = '台' * 5000
data = text.encode('utf-7-imap')
f = io.TextIOWrapper(io.BytesIO(data), encoding='utf-7-imap')
print(f.read(100) == text[:100])

UnicodeDecodeError: 'utf-7-imap' codec can't decode bytes in position 0-8191: unterminated shift sequence

Expected: True. Doc/library/codecs.rst says the joined output of calls to the incremental decode method is the same as decoding the joined input with the stateless decoder.

The incremental encoder breaks the same sentence more mildly: encoding '~peter/mail/台北/日本語' one character at a time joins to b'~peter/mail/&U,A-&Uxc-/&ZeU-&Zyw-&ip4-' where one shot gives b'~peter/mail/&U,BTFw-/&ZeVnLIqe-', though both decode back to the same text.

The codec is new in 3.16, so no released version is affected.

CPython versions tested on:

CPython main branch

Operating systems tested on:

Linux

Linked PRs

• gh-156967

Start with the utf-7-imap incremental decoder and stream reader paths exercised by io.TextIOWrapper.read, codecs.iterdecode, and codecs.StreamReader.read. Read the incremental codec guidance in Doc/library/codecs.rst, then verify that split input works and that joined incremental results match the stateless decoder behavior.