[BUGFIX] Decode tag attributes during HTML parsing

When parsing HTML in parsefunc, we need to make sure
that the attributes that are extracted for further processing
are properly HTML decoded, as per definition they are encoded in
the HTML being parsed.

At the same time we revert a decode that was applied in ExternalUrlLinkBuilder,
as this is the wrong place to handle HTML decoding, as the link builder
is as well called with data from other contexts (TypoScript, data from link fields)
that must not have the URL HTML encoded.

This fix is required to allow email links to add subject and body arguments like:
test@foo.test?subject=foo%20bar&body=baz

and not have the ampersand in the url double encoded when generating links
from RTE content.

Additionally we need to properly encode the masked value for the JavaScript
URI, to prevent browsers from url decoding wrong values.

Releases: master, 9.5
Resolves: #89264
Related: #81625
Related: #80985
Change-Id: I3be2478df83cffc0dfe6af918eb224b4ee480b10
Reviewed-on: https://review.typo3.org/c/Packages/TYPO3.CMS/+/61799


Tested-by: TYPO3com <noreply@typo3.com>
Tested-by: Daniel Goerz <daniel.goerz@posteo.de>
Tested-by: Benni Mack <benni@typo3.org>
Reviewed-by: Daniel Goerz <daniel.goerz@posteo.de>
Reviewed-by: Benni Mack <benni@typo3.org>