Previously TagExtractor.parse_text would return all matches, including
duplicates, and would therefore potentially link a tag to a set or
sample multiple times (if used mutiple times in the info field).

This commit ensures tags are extracted only once, but will not affect
the link in the frontend