Parser: Allow literal control chars in logfmt decoder

[Timbus]

What this PR does / why we need it:
This allows logfmt to handle newlines so long as they are quoted as expected. It still unquotes control chars so nothing should ever break. It just throws less errors.

Rant
There are other issues I have ignored. The main one being that the ScanKeyval parser tries to be more efficient by avoiding calling unquoteBytes unless it sees a \, but it relies on unquoteBytes to handle utf8. So a unicode token without a \ may not be decoded correctly.
unquoteBytes also has its own logic to check for slashes so it's inefficient
unquoteBytes also seems to expect that the given string starts and ends with " but surely anything passing the string has already checked for these delimiters so, why pass them?.. another mystery. And also why does it care if it finds an unescaped quote char, if its not a stream scanner?
Basically someone should inline some of unquoteBytes code into ScanKeyval, is what I'm getting at.

Which issue(s) this PR fixes:
Fixes #3927

Checklist

• Documentation added
• Tests updated

[CLAassistant]

All committers have signed the CLA.

[cyriltovena]

LGTM