Using IndexRange in FlatFunctor

[pgrete]

          yeah I remember cuda builds had given me trouble with that. This [one](https://godbolt.org/z/4W37ejM7E) uses some more template helper structs to also handle `IndexRange`s in the constructor. I imagine you could use similar things to handle all the `par_dispatch` overloads as well

Originally posted by @acreyes in #1130 (comment)

As discussed above, there might be some general enhancement around this.
Opening this issue with the reference to the comment (and the two above that one) as it included some test code.