Have a look at the document() function documentation.
You can use document() to load further XML documents during the transformation process. They are loaded as node sets. That means you would initially feed the XML that contains the file names to load to the XSLT, and take it from there: