[llvm] Delete the incorrect assert that assumes size of `APFloat` is the same as `IEEEFloat`. (PR #111780)

[Nikita Popov via llvm-commits]

nikic wrote:

Yeah, I don't think we should change the assert upstream. When you add support for your new floating point type upstream, we can decide whether we're willing to increase the APFloat size for it, or you need to reduce the size of your representation (e.g. by indirecting via an allocation, just like DoubleAPFloat does).

https://github.com/llvm/llvm-project/pull/111780