And of course it is not possible to write a Swift compiler in Swift before the language exists.
So no, not odd at all. Quite logical actually.
No, but you can do a self-hosting port after you have a first rudimentary compiler. Many compilers have been built like this. Some did it early in language development with a purpouse built one-off bootstrapping compiler, some did it after the language was somewhat stable (eg Go).
ObWP: https://en.wikipedia.org/wiki/Bootstrapping_%28compilers%29
Aside from the value of maintaining LLVM as a shared codebase, the concern was that writing a compiler in Swift could bias early language design tradeoffs towards decisions that would make it easier to write a compiler at the expense of the larger universe of potential applications.
From a language design point of view, the things needed to get a language to the point where it could successfully and efficiently compile itself would skew or re-order priorities in a much different fashion than if they were to focus on building and iterating on an applications and systems language.
tl;dr: LLVM and Clang and their assorted toolsets do quite a bit of valuable work; reinventing that to serve the language they're building is a feedback loop is a downside that also removes some upside from the equation, too.
That being said, C++ is probably a better choice, because dynamic dispatch sucks and there's better platform support.