support sets

[J-Zeitler]

considerations:

• this implementation clashes a bit with slicez.Set()
• current implementation follows rest of the convention in the lib, could instead/also expose methods on the generic Set type

[crholm]

Not against it, slicez.Set() is map[E comparable]bool to have a map behave like a Set.

For performance reasons the setz package seems resonable, supporting mutations. But is the bool vs struct that much different or it the reason looking for eg effective Intersection, Difference and so on operations? If so, that might be a good reason to add them on a key level on in mapz instead?