Allow "void" as a generic type parameter.

[lrhn]

The void type is only allowed as a function return type. It is equivalent to "Null" except for documentation purposes (and allowing "return;" statements in the body).
The documentation meaning is that the function does not return a meaningful value, and its result shouldn't be used.

For Future, we have the same documentation need. Some Futures carry no meaningful value, they are effectively the delayed results of evaluating a void function.
It would improve documentation if these futures could be typed as "Future<void>". By duality, it would be equivalent to "Future<Null>", but documenting a different intent.

It's likely to also be useful in other places. Perhaps we should allow "void" everywhere we currently allow "dynamic".

[gbracha]

I really don't see the point in letting void spread its malaise any further. As a type it means nothing - it has no values. Future is the only case here, and you can write Future<Null> if you want now, which accurately describes what the Future is carrying.

---

Set owner to @gbracha.
Added Accepted label.

[munificent]

Duplicate of #27504. (Yes, this one is much older, but that one is linked to by more things.)