`-irrb` Sometimes returns corrupted base64 body

[nil0x42]

httpx version:

1.2.6

Current Behavior:

Sometimes, binary body is altered, and decoding base64 body returns unexpected data (who looks to have been pre-encoded as utf-8)

Here is a POC:

$ echo 'https://zc-console.taobao.com/favicon.ico' | httpx -silent -json -irrb | jq .body -j | base64 -d | wc -c

2246

$ curl -s 'https://zc-console.taobao.com/favicon.ico' | wc -c

1863

Note that in this specific case, Content-Type header is set to image/x-icon;charset=GBK, so underlying http library might be automatically trying to encode data, given that charset is specified.

Expected Behavior:

-irrb should always returns a base64 version of the raw, original body's binary content, with no previous encoding

Anything else:

related to #813, #814