introduce a generic `closed_under` definition

[affeldt-aist]

as explained in the following comment by @t6s

#1868 (review)

and completed by the notation proposed by @CohenCyril in this other comment

#1868 (comment)