The codecs module mostly but not fully supports non str<->bytes codecs

Bug report

Bug description:

There's in the codecs module an option for charbuffertype which allows bytes<->bytes codecs, but if there were a type for the encoded value, it would allow cleaner implementation of str<->str codecs, or more...
Most codecs in CPython use decoded str and encoded bytes, there are also codecs that have decoded bytes and encoded bytes, like base64 (note that a base64 codec arguably could be one that considers decoded bytes and encoded str).

I had been more or less perverting the codecs module to implement things like COBS (fundamentally bytes<->bytes), bubblebabble (I did bytes<->bytes but it should encode to str), and recently ANSI escape filtering, as a str<->str codec that works by implementing a virtual terminal, with a screen buffer. To make it work I only had to do self.bytebuffer = "" in the StreamReader constructor, but since there's a charbuffertype, maybe we could use a bytesbuffertype, and open the flood gates to all kinds of creative abuse.

CPython versions tested on:

3.12

Operating systems tested on:

Linux

从 codecs 模块开始，检查 StreamReader 构造函数，包括现有的 charbuffertype 行为。确定编码值类型和解码值类型的预期模型，然后定义超出 str/bytes 的更广泛 codec 支持应涵盖哪些内容；只有在该范围及其预期行为确定并实现后，工作才算完成。