Description

In infer_content_encoding, we set the fallback encoding to the UTF-8 standard instead of the (very) legacy Latin-1 encoding.

As a consequence, get_text() outputs (and thus HAR dumps) may differ from previous version, advocating for more than a patch release.

(Note that to reach full coverage I had to perform a mini-refact in mitmproxy.io.har -> extract a little untested snippet into an unitary testable function)

Checklist

• I have updated tests where applicable.
• I have added an entry to the CHANGELOG.