Not all plain text-based mimetypes starts with text/:
i.e. application/sql for SQL dumps (already handled in FileExcludeFilters),
or application/postscript for PS images (also handled by PostscriptDSCExtractor).
However, there are most likely to be more.
This patch aims at handling the issue in more generic way:
it just skips PlaintextExtractor for any large file, utilizing extractor metadata introduced in D19109: [Extractor] Add metadata to extractors.
Alternative solution would be using QMimeType::inherits instead.