Normal file page links can be very long (up to ~300 characters) so there is need for a shorter URL to be exposed via the imageinfo API (see T119686 for a use case). One possibility would be to expose the page ID, and then an URL could be constructed as scriptUrl + '?curid=' + pageId (script URL is already exposed by the filerepoinfo API).
This will be a bit hacky since shared-DB file repositories are not technically required to have file pages, but we already make the same assumptions with the description page URL.