Each paginated page should be a real, crawlable URL with a self-referencing canonical and ordinary links to the next and previous pages. Canonicalising every page to page one is a frequent and damaging mistake: it tells the engine that pages two onwards do not exist, and the products only linked from them disappear from discovery.
Infinite scroll and "load more" buttons are the same problem in a nicer costume. If the additional items only appear after a JavaScript event, crawlers never see them. Back any such interface with paginated URLs that work without JavaScript, which also helps users who want to share or bookmark a position in the list.