Lack of language improvement. Sure, you can do anything with macros, but then when you open some random open source library it may be using different macros (or macros that are named the same way but act differently) than those that you use. Putting something in the language forces everybody to use the same standard. It can be good and it can be bad.
There's a middle-ground stance that's pretty common in the Elixir ecosystem: five or six libraries get written to solve a problem over the course of a year or two, people play around with them, and then a common set of primitives gets factored out of those libraries and made into an interchange library that all those libs support.
That interchange library can then be integrated into the Elixir stdlib—that's what happened with DateTimes in Elixir—but there's no reason it needs to be. It works just as well to just keep the interchange-type library as its own tiny little dependency that all these libraries pull in.
This is one of the awesome things about Plug, in the sense that if your library or what have you targets Plug it'll work reasonably well (if not entirely seamlessly) in a Phoenix app, a Sugar app, a pure-Plug app, a $SomeOtherFramework app, or what have you.
Plug happens to be somewhere in between the two outcomes you've presented, of course; its development happens very closely to that of Elixir (and it's arguably a sub-project thereof), but it ain't a core part of the language itself (last I checked; maybe this has changed in recent years).