Unmarshalling of links when transforming Editor XHTML store CDATA when there are non-latin characters

Steps to reproduce:

1. Create a user that has a non latin character in his name, e.g. John Stærk
2. Navigate to a confluence page
3. Create a new comment
4. Enter as the data a mention to John Stærk
5. Set a breakpoint on line 122 of EditorLinkBodyUnmarshaller, i.e.

   if (StringUtils.equals(defaultAlias, linkBodyString))

6. Click Save

Expected Outcome:
The default alias that contains the plain text representation of John Stærk should match the unicode representation, John Stærk. Either that, or the default alias should be the unicode representation at this point, resulting in the current representation matching.

Actual Outcome:
A blind comparison is done, which ignores whether it is unicode altogether, which interprets the values as different, and therefore determines that the body text should be stored (unnecessarily, as the default alias provides the necessary information)

This has caused a bug to surface in Confluence Dashboard All Updates CONF-28169. This does not completely fix CONF-28169 as it only addresses the ID/label variance, not whether it should be an ID or a label. More details are in STRM-2133 which was incorrectly linked.